Welcome to swiss_army_keras’s documentation!

Introduction

A library to help with the development of AI models with Keras, with a focus on edge / IoT applications.

Motivation

Building CNN experiments with flexible and scalable augmentation pipelines, with the ability to cobine different encoder and decoder architectures, can require a huge initial effort. This is even more true in the case of edge-AI models, where models need to be build taking into account more severe constraints and quantization is usually required.

This library wants you to focus on dataset, model architecture and hyperparameters tuning, without worring about the rest.

It provides several helper classes which help in the development of CNN based AI models for edge IoT applications, where resources are limited and model quantization is reccomended.

Features

The main features of the library are the following:

1. Flexible, efficient and scalable Dataset management with augmentation pipelines leveraging albumentations (https://albumentations.ai/) and td-data (https://www.tensorflow.org/guide/data)

2. Training driver with builtin callbacks, configurable backbone unfreezing, and quantized model generation

3. Helper classes to easiliy combine pretrained backbones for Edge AI applications with the desired segmentation and classification architectures

4. Additional loss functions and optimizers which are not part of the Keras distribution, as for now

Examples

You can find here sample notebooks which show the usage of varius feature of the library in a production like experiment and setup.

1. Classification example: a classification experiment based on distiller classifier layer, with MobilenetV3Plus minimalistic feature extractor backbone.

2. Segmentation example: a segmentation experiment based on DeepLabv3+ segmentation decoder with MobilenetV3Plus minimalistic encoder.

3. Model User Guide. A notebook with documentation about the different possible models provided by this library

Classification Example

import os
import sys
from base64 import b64encode

import pathlib

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from functools import partial
import albumentations as albu

Matplotlib created a temporary config/cache directory at /tmp/matplotlib-14mjotfm because the default path (/.config/matplotlib) is not a writable directory; it is highly recommended to set the MPLCONFIGDIR environment variable to a writable directory, in particular to speed up the import of Matplotlib and to better support multiprocessing.

Require 24 GB of GPU RAM

gpus = tf.config.experimental.list_physical_devices("GPU")
#os.environ['TF_GPU_ALLOCATOR'] = 'cuda_malloc_async'


if gpus:
    try:
        tf.config.experimental.set_virtual_device_configuration(
            gpus[0],
            [
                tf.config.experimental.VirtualDeviceConfiguration(
                    memory_limit=1024 * 24
                )
            ],
        )
    except RuntimeError as e:
        print(e)

Download and setup dataset

flowers_root = tf.keras.utils.get_file(
    'flower_photos',
    'https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz',
    untar=True)
flowers_root = pathlib.Path(flowers_root)

Import swiss_army keras lib

#Uncomment to install dependencies, if needed
#!pip3 install --upgrade git+https://github.com/waterviewsrl/swiss-army-keras.git
#!pip3 install typeguard
#!pip3 install --upgrade irondomo
#!pip3 install --upgrade pillow
#!pip3 install --upgrade adabelief_tf
from swiss_army_keras.dataset_utils import ClassificationAlbumentationsDataLoader

from swiss_army_keras import models, base, utils, losses, optimizers

Define input tensor and build classifier model with default MobileNetV3Large minimalistic backbone

This model (distiller) defines a top classification layer which tries to go beyond the usual GlobalPooling + Dense layer, by preserving information about spatial relation of macrofeatures in the last convolutional layer output.

inp = tf.keras.layers.Input(shape=(512, 512, 3), batch_size=1)

model = models.distiller_classifier(inp, 5, macrofeatures_number=5, size=256)

WARNING:tensorflow:`input_shape` is undefined or non-square, or `rows` is not 224. Weights for input shape (224, 224) will be loaded as the default.

Define augmentation pipeline (see https://albumentations.ai/docs/)

def get_training_augmentation(width=512, height=512):
    train_transform = [


        #albu.RandomRotate90(p=0.5),

        #albu.OneOf([
        #
        #
        #], p=1),
        #albu.Resize(width, height, 2, p=1),

        #albu.ShiftScaleRotate(shift_limit = 0.05, scale_limit=(0.1, 0.1), interpolation=2, rotate_limit=25, p=0.3),

        albu.HorizontalFlip(p=0.5),
        #albu.VerticalFlip(p=0.15),

        albu.Compose([
        albu.OneOf([
            albu.ISONoise(p=1),
            albu.MultiplicativeNoise(p=1),
            albu.GaussNoise(p=1),
            albu.ImageCompression(p=1),

            ],
            p=0.2,
        ),

        #albu.Perspective(pad_mode=2, p=0.1),

        albu.OneOf(
            [
                albu.CLAHE(p=1),
                albu.RandomBrightnessContrast(p=1),
                albu.Sharpen(p=1),
                albu.Emboss(p=1),
                albu.RandomGamma(p=1),
            ],
            p=0.5,
        ),

        albu.OneOf(
            [
                albu.Blur(blur_limit=5, p=1),
                albu.MedianBlur(blur_limit=5, p=1),
                albu.MotionBlur(blur_limit=5, p=1),
            ],
            p=0.25,
        ),

        albu.OneOf(
            [
                albu.RandomBrightnessContrast(p=1),
                albu.HueSaturationValue(p=1),
                albu.RGBShift(p=1),
            ],
            p=0.25,
        ),

        albu.OneOf(
            [
                #albu.ElasticTransform(alpha=10, sigma=10 * 0.05, alpha_affine=10 * 0.03, p=0.5),
                albu.GridDistortion(p=0.5),
                albu.OpticalDistortion(distort_limit=0.3, shift_limit=0.05, p=1)
            ],
            p=0.25,
        )], p=0.5),

        #DropsOnLens(maxR=40, minR=30, p=0.1),
        albu.ToGray(p=0.2)
    ]

    return albu.Compose(train_transform)
    #return albu.Compose([])

Initialize classification dataloader

This will create a number of workers equal to half the number of CPU cores on the machine. These workers will produce random augmentations according to the defined pipeline, so that at each epoch every batch will change.

Please note: Passing normalization=model.preprocessing grants that the correct normalization is applied in a consistent way during training, evaluation and quantization.

Please note: This will produce a lot of output, depending also on the number of CPUs, to monitor workers connectivity and liveliness

dl = ClassificationAlbumentationsDataLoader(str(flowers_root), width=512, height=512, train_augmentations=get_training_augmentation(), batch_size=16, normalization=model.preprocessing)

WARNING:root:Normalizing in range: [  0 255]

Class label daisy will have id: 0
Class label dandelion will have id: 1
Class label roses will have id: 2
Class label sunflowers will have id: 3
Class label tulips will have id: 4
train_val_test_split: [2936, 367, 367]

WARNING:root:KEYS: None
WARNING:root:Router on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:KEYS: (b'P+S690P{iVPfx<aFJwxfSY^ugFzjuWOnaIh!o7J<', b':$80ST.hxA5xL7c+@$3YTEohOR^GhrJ2$qzN@bR^')
WARNING:root:KEYS: None
WARNING:root:Attaching HOST CURVE keys.
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 128, 'value': 96, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CLOSED'}
WARNING:root:Router on "ipc:///tmp/swiss_army_keras_classification_curve_socket_QbuLn4Q="
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4, 'value': 297, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECT_RETRIED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 128, 'value': 97, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CLOSED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 8, 'value': 107, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_LISTENING'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4, 'value': 297, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECT_RETRIED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_curve_socket_QbuLn4Q=): {'event': 8, 'value': 108, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_curve_socket_QbuLn4Q=', 'description': 'EVENT_LISTENING'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 98, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 109, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 99, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 115, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 100, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 116, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 101, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 112, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 121, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 103, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 96, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 125, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 104, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 126, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 97, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 127, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 105, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 106, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 110, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 111, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 107, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 129, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 108, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 131, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 109, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 130, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 110, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 120, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 113, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 112, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 114, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 112, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 117, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 112, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 114, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 137, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 1, 'value': 113, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 32, 'value': 134, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_classification_clear_socket_QbuLn4Q=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}

Define train val and test sets

train_set, val_set, test_set = dl.build_datasets()

View a batch of augmentetd images from train set

dl.show_images(16)

from adabelief_tf import AdaBeliefOptimizer
from swiss_army_keras.training_utils import TrainingDriver

accuracy = tf.keras.metrics.CategoricalAccuracy()

Train for 40 eochs (20 with frozen backend + 20 with unfrozen backend) with the desired optimizer, loss and metrices.

This will also produce h5 model, f16 and ui8 quantized models, metrics and loss history graphs, all named according to the name (prova_multi in this case) + startup datetime.

Please Note: TrainingDriver implicitly defines callbacks for ModelCheckpoint and for EarlyStopping with 15 epochs patience. The saved models, quantized and not, corresponds to the best validation loss found during training.

td = TrainingDriver(model, 'prova_multi', AdaBeliefOptimizer(learning_rate=1e-4, epsilon=1e-14, rectify=True), tf.keras.losses.CategoricalCrossentropy(label_smoothing=0.1), [accuracy], train_set, val_set, test_set, 20 )
td.run()

Please check your arguments if you have upgraded adabelief-tf from version 0.0.1.
Modifications to default arguments:
                           eps  weight_decouple    rectify
-----------------------  -----  -----------------  -------------
adabelief-tf=0.0.1       1e-08  Not supported      Not supported
>=0.1.0 (Current 0.2.1)  1e-14  supported          default: True
SGD better than Adam (e.g. CNN for Image Classification)    Adam better than SGD (e.g. Transformer, GAN)
----------------------------------------------------------  ----------------------------------------------
Recommended epsilon = 1e-7                                  Recommended epsilon = 1e-14
For a complete table of recommended hyperparameters, see
https://github.com/juntang-zhuang/Adabelief-Optimizer
You can disable the log message by setting "print_change_log = False", though it is recommended to keep as a reminder.

Model: "model"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to
==================================================================================================
input_1 (InputLayer)            [(1, 512, 512, 3)]   0
__________________________________________________________________________________________________
MobileNetV3Large_backbone (Func [(1, 256, 256, 64),  1437608     input_1[0][0]
__________________________________________________________________________________________________
average_pooling2d (AveragePooli (1, 5, 5, 960)       0           MobileNetV3Large_backbone[0][4]
__________________________________________________________________________________________________
depthwise_conv2d (DepthwiseConv (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_1 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_2 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_3 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_4 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
flatten (Flatten)               (1, 960)             0           depthwise_conv2d[0][0]
__________________________________________________________________________________________________
flatten_1 (Flatten)             (1, 960)             0           depthwise_conv2d_1[0][0]
__________________________________________________________________________________________________
flatten_2 (Flatten)             (1, 960)             0           depthwise_conv2d_2[0][0]
__________________________________________________________________________________________________
flatten_3 (Flatten)             (1, 960)             0           depthwise_conv2d_3[0][0]
__________________________________________________________________________________________________
flatten_4 (Flatten)             (1, 960)             0           depthwise_conv2d_4[0][0]
__________________________________________________________________________________________________
dropout (Dropout)               (1, 960)             0           flatten[0][0]
__________________________________________________________________________________________________
dropout_1 (Dropout)             (1, 960)             0           flatten_1[0][0]
__________________________________________________________________________________________________
dropout_2 (Dropout)             (1, 960)             0           flatten_2[0][0]
__________________________________________________________________________________________________
dropout_3 (Dropout)             (1, 960)             0           flatten_3[0][0]
__________________________________________________________________________________________________
dropout_4 (Dropout)             (1, 960)             0           flatten_4[0][0]
__________________________________________________________________________________________________
concatenate (Concatenate)       (1, 4800)            0           dropout[0][0]
                                                                 dropout_1[0][0]
                                                                 dropout_2[0][0]
                                                                 dropout_3[0][0]
                                                                 dropout_4[0][0]
__________________________________________________________________________________________________
dense (Dense)                   (1, 256)             1229056     concatenate[0][0]
__________________________________________________________________________________________________
dropout_5 (Dropout)             (1, 256)             0           dense[0][0]
__________________________________________________________________________________________________
dense_1 (Dense)                 (1, 5)               1285        dropout_5[0][0]
==================================================================================================
Total params: 2,792,749
Trainable params: 1,355,141
Non-trainable params: 1,437,608
__________________________________________________________________________________________________
Epoch 1/20
WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

/usr/local/lib/python3.6/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
  category=CustomMaskWarning)
WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

183/183 [==============================] - ETA: 0s - loss: 2.0143 - categorical_accuracy: 0.5249WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

183/183 [==============================] - 16s 55ms/step - loss: 2.0143 - categorical_accuracy: 0.5249 - val_loss: 1.8842 - val_categorical_accuracy: 0.6591
Epoch 2/20
183/183 [==============================] - 10s 50ms/step - loss: 1.6382 - categorical_accuracy: 0.7510 - val_loss: 1.4155 - val_categorical_accuracy: 0.8040
Epoch 3/20
183/183 [==============================] - 10s 52ms/step - loss: 1.2311 - categorical_accuracy: 0.8463 - val_loss: 1.1303 - val_categorical_accuracy: 0.8438
Epoch 4/20
183/183 [==============================] - 10s 51ms/step - loss: 1.0534 - categorical_accuracy: 0.8736 - val_loss: 1.0248 - val_categorical_accuracy: 0.8551
Epoch 5/20
183/183 [==============================] - 10s 54ms/step - loss: 0.9722 - categorical_accuracy: 0.8863 - val_loss: 0.9528 - val_categorical_accuracy: 0.8693
Epoch 6/20
183/183 [==============================] - 10s 53ms/step - loss: 0.9048 - categorical_accuracy: 0.8975 - val_loss: 0.8957 - val_categorical_accuracy: 0.8807
Epoch 7/20
183/183 [==============================] - 10s 51ms/step - loss: 0.8534 - categorical_accuracy: 0.9088 - val_loss: 0.8701 - val_categorical_accuracy: 0.8778
Epoch 8/20
183/183 [==============================] - 10s 51ms/step - loss: 0.8143 - categorical_accuracy: 0.9208 - val_loss: 0.8423 - val_categorical_accuracy: 0.8778
Epoch 9/20
183/183 [==============================] - 10s 53ms/step - loss: 0.7829 - categorical_accuracy: 0.9255 - val_loss: 0.8147 - val_categorical_accuracy: 0.8864
Epoch 10/20
183/183 [==============================] - 10s 53ms/step - loss: 0.7551 - categorical_accuracy: 0.9324 - val_loss: 0.7910 - val_categorical_accuracy: 0.8977
Epoch 11/20
183/183 [==============================] - 10s 53ms/step - loss: 0.7279 - categorical_accuracy: 0.9372 - val_loss: 0.7781 - val_categorical_accuracy: 0.8977
Epoch 12/20
183/183 [==============================] - 10s 53ms/step - loss: 0.7150 - categorical_accuracy: 0.9430 - val_loss: 0.7659 - val_categorical_accuracy: 0.9006
Epoch 13/20
183/183 [==============================] - 10s 53ms/step - loss: 0.6990 - categorical_accuracy: 0.9406 - val_loss: 0.7563 - val_categorical_accuracy: 0.9006
Epoch 14/20
183/183 [==============================] - 10s 54ms/step - loss: 0.6879 - categorical_accuracy: 0.9423 - val_loss: 0.7415 - val_categorical_accuracy: 0.8977
Epoch 15/20
183/183 [==============================] - 10s 52ms/step - loss: 0.6765 - categorical_accuracy: 0.9491 - val_loss: 0.7342 - val_categorical_accuracy: 0.8977
Epoch 16/20
183/183 [==============================] - 10s 52ms/step - loss: 0.6601 - categorical_accuracy: 0.9529 - val_loss: 0.7322 - val_categorical_accuracy: 0.8977
Epoch 17/20
183/183 [==============================] - 10s 52ms/step - loss: 0.6540 - categorical_accuracy: 0.9532 - val_loss: 0.7276 - val_categorical_accuracy: 0.8920
Epoch 18/20
183/183 [==============================] - 10s 54ms/step - loss: 0.6433 - categorical_accuracy: 0.9583 - val_loss: 0.7173 - val_categorical_accuracy: 0.9091
Epoch 19/20
183/183 [==============================] - 10s 52ms/step - loss: 0.6379 - categorical_accuracy: 0.9577 - val_loss: 0.7139 - val_categorical_accuracy: 0.9034
Epoch 20/20
183/183 [==============================] - 9s 50ms/step - loss: 0.6310 - categorical_accuracy: 0.9604 - val_loss: 0.7095 - val_categorical_accuracy: 0.9062

WARNING:root:Unfreezing Model

Model: "model"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to
==================================================================================================
input_1 (InputLayer)            [(1, 512, 512, 3)]   0
__________________________________________________________________________________________________
MobileNetV3Large_backbone (Func [(1, 256, 256, 64),  1437608     input_1[0][0]
__________________________________________________________________________________________________
average_pooling2d (AveragePooli (1, 5, 5, 960)       0           MobileNetV3Large_backbone[0][4]
__________________________________________________________________________________________________
depthwise_conv2d (DepthwiseConv (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_1 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_2 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_3 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_4 (DepthwiseCo (1, 1, 1, 960)       24960       average_pooling2d[0][0]
__________________________________________________________________________________________________
flatten (Flatten)               (1, 960)             0           depthwise_conv2d[0][0]
__________________________________________________________________________________________________
flatten_1 (Flatten)             (1, 960)             0           depthwise_conv2d_1[0][0]
__________________________________________________________________________________________________
flatten_2 (Flatten)             (1, 960)             0           depthwise_conv2d_2[0][0]
__________________________________________________________________________________________________
flatten_3 (Flatten)             (1, 960)             0           depthwise_conv2d_3[0][0]
__________________________________________________________________________________________________
flatten_4 (Flatten)             (1, 960)             0           depthwise_conv2d_4[0][0]
__________________________________________________________________________________________________
dropout (Dropout)               (1, 960)             0           flatten[0][0]
__________________________________________________________________________________________________
dropout_1 (Dropout)             (1, 960)             0           flatten_1[0][0]
__________________________________________________________________________________________________
dropout_2 (Dropout)             (1, 960)             0           flatten_2[0][0]
__________________________________________________________________________________________________
dropout_3 (Dropout)             (1, 960)             0           flatten_3[0][0]
__________________________________________________________________________________________________
dropout_4 (Dropout)             (1, 960)             0           flatten_4[0][0]
__________________________________________________________________________________________________
concatenate (Concatenate)       (1, 4800)            0           dropout[0][0]
                                                                 dropout_1[0][0]
                                                                 dropout_2[0][0]
                                                                 dropout_3[0][0]
                                                                 dropout_4[0][0]
__________________________________________________________________________________________________
dense (Dense)                   (1, 256)             1229056     concatenate[0][0]
__________________________________________________________________________________________________
dropout_5 (Dropout)             (1, 256)             0           dense[0][0]
__________________________________________________________________________________________________
dense_1 (Dense)                 (1, 5)               1285        dropout_5[0][0]
==================================================================================================
Total params: 2,792,749
Trainable params: 2,768,349
Non-trainable params: 24,400
__________________________________________________________________________________________________
Epoch 1/20
WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

183/183 [==============================] - ETA: 0s - loss: 0.7263 - categorical_accuracy: 0.9095WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

WARNING:tensorflow:Model was constructed with shape (1, 512, 512, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 512, 512, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 512, 512, 3).

183/183 [==============================] - 63s 244ms/step - loss: 0.7263 - categorical_accuracy: 0.9095 - val_loss: 0.8143 - val_categorical_accuracy: 0.8608
Epoch 2/20
183/183 [==============================] - 42s 226ms/step - loss: 0.6166 - categorical_accuracy: 0.9611 - val_loss: 0.7439 - val_categorical_accuracy: 0.8807
Epoch 3/20
183/183 [==============================] - 41s 223ms/step - loss: 0.5775 - categorical_accuracy: 0.9775 - val_loss: 0.6824 - val_categorical_accuracy: 0.9062
Epoch 4/20
183/183 [==============================] - 41s 224ms/step - loss: 0.5598 - categorical_accuracy: 0.9843 - val_loss: 0.6625 - val_categorical_accuracy: 0.9091
Epoch 5/20
183/183 [==============================] - 41s 224ms/step - loss: 0.5425 - categorical_accuracy: 0.9870 - val_loss: 0.6520 - val_categorical_accuracy: 0.9006
Epoch 6/20
183/183 [==============================] - 41s 225ms/step - loss: 0.5324 - categorical_accuracy: 0.9908 - val_loss: 0.6757 - val_categorical_accuracy: 0.8892
Epoch 7/20
183/183 [==============================] - 42s 228ms/step - loss: 0.5218 - categorical_accuracy: 0.9911 - val_loss: 0.6407 - val_categorical_accuracy: 0.9176
Epoch 8/20
183/183 [==============================] - 41s 223ms/step - loss: 0.5139 - categorical_accuracy: 0.9928 - val_loss: 0.6174 - val_categorical_accuracy: 0.9318
Epoch 9/20
183/183 [==============================] - 42s 225ms/step - loss: 0.5096 - categorical_accuracy: 0.9915 - val_loss: 0.6094 - val_categorical_accuracy: 0.9261
Epoch 10/20
183/183 [==============================] - 42s 227ms/step - loss: 0.4980 - categorical_accuracy: 0.9949 - val_loss: 0.6024 - val_categorical_accuracy: 0.9318
Epoch 11/20
183/183 [==============================] - 42s 228ms/step - loss: 0.4962 - categorical_accuracy: 0.9928 - val_loss: 0.6056 - val_categorical_accuracy: 0.9290
Epoch 12/20
183/183 [==============================] - 42s 226ms/step - loss: 0.4871 - categorical_accuracy: 0.9956 - val_loss: 0.6182 - val_categorical_accuracy: 0.9318
Epoch 13/20
183/183 [==============================] - 42s 225ms/step - loss: 0.4838 - categorical_accuracy: 0.9949 - val_loss: 0.5962 - val_categorical_accuracy: 0.9318
Epoch 14/20
183/183 [==============================] - 41s 223ms/step - loss: 0.4812 - categorical_accuracy: 0.9945 - val_loss: 0.5968 - val_categorical_accuracy: 0.9375
Epoch 15/20
183/183 [==============================] - 41s 222ms/step - loss: 0.4762 - categorical_accuracy: 0.9959 - val_loss: 0.5996 - val_categorical_accuracy: 0.9290
Epoch 16/20
183/183 [==============================] - 41s 224ms/step - loss: 0.4726 - categorical_accuracy: 0.9956 - val_loss: 0.5892 - val_categorical_accuracy: 0.9347
Epoch 17/20
183/183 [==============================] - 41s 224ms/step - loss: 0.4675 - categorical_accuracy: 0.9973 - val_loss: 0.5873 - val_categorical_accuracy: 0.9318
Epoch 18/20
183/183 [==============================] - 41s 224ms/step - loss: 0.4639 - categorical_accuracy: 0.9966 - val_loss: 0.5792 - val_categorical_accuracy: 0.9318
Epoch 19/20
183/183 [==============================] - 42s 225ms/step - loss: 0.4599 - categorical_accuracy: 0.9980 - val_loss: 0.5845 - val_categorical_accuracy: 0.9318
Epoch 20/20
183/183 [==============================] - 41s 222ms/step - loss: 0.4561 - categorical_accuracy: 0.9973 - val_loss: 0.5657 - val_categorical_accuracy: 0.9403

WARNING:root:Quantizing Model

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

INFO:tensorflow:Assets written to: /tmp/tmp2chah2d2/assets

INFO:tensorflow:Assets written to: /tmp/tmp2chah2d2/assets
/usr/local/lib/python3.6/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
  category=CustomMaskWarning)

WARNING:tensorflow:Issue encountered when serializing table_initializer.
Type is unsupported, or the types of the items don't match field type in CollectionDef. Note this is a warning and probably safe to ignore.
'NoneType' object has no attribute 'name'

WARNING:tensorflow:Issue encountered when serializing table_initializer.
Type is unsupported, or the types of the items don't match field type in CollectionDef. Note this is a warning and probably safe to ignore.
'NoneType' object has no attribute 'name'
WARNING:absl:For model inputs containing unsupported operations which cannot be quantized, the `inference_input_type` attribute will default to the original type.

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

INFO:tensorflow:Assets written to: /tmp/tmpjiq88zc4/assets

INFO:tensorflow:Assets written to: /tmp/tmpjiq88zc4/assets
/usr/local/lib/python3.6/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
  category=CustomMaskWarning)

WARNING:tensorflow:Issue encountered when serializing table_initializer.
Type is unsupported, or the types of the items don't match field type in CollectionDef. Note this is a warning and probably safe to ignore.
'NoneType' object has no attribute 'name'

WARNING:tensorflow:Issue encountered when serializing table_initializer.
Type is unsupported, or the types of the items don't match field type in CollectionDef. Note this is a warning and probably safe to ignore.
'NoneType' object has no attribute 'name'

<Figure size 432x288 with 0 Axes>

Show some reuslts with corresponding reference classification

dl.show_results(model, 16)

<Figure size 1584x1584 with 0 Axes>

Example segmentation with DeepLabv3+

import numpy as np
from glob import glob
import tensorflow as tf

from tensorflow import keras
from tensorflow.keras.layers import Input, Conv2D, Dropout, Activation, UpSampling2D, GlobalMaxPooling2D, multiply
from tensorflow.keras.backend import max

gpus = tf.config.experimental.list_physical_devices("GPU")
#os.environ['TF_GPU_ALLOCATOR'] = 'cuda_malloc_async'


if gpus:
    try:
        tf.config.experimental.set_virtual_device_configuration(
            gpus[0],
            [
                tf.config.experimental.VirtualDeviceConfiguration(
                    memory_limit=1024 * 24
                )
            ],
        )
    except RuntimeError as e:
        print(e)

This example requires swiss_army_keras or higher:

pip3 install swiss_army_keras

#!pip3 install --upgrade git+https://github.com/waterviewsrl/swiss-army-keras.git
from swiss_army_keras import models, base, utils, quantization_utils, training_utils, dataset_utils

Matplotlib created a temporary config/cache directory at /tmp/matplotlib-g1cfk7fj because the default path (/.config/matplotlib) is not a writable directory; it is highly recommended to set the MPLCONFIGDIR environment variable to a writable directory, in particular to speed up the import of Matplotlib and to better support multiprocessing.

# the indicator of a fresh run
first_time_running = False

# user-specified working directory
filepath = './oxford_iiit/'

The Oxford-IIIT Pets dataset

This example applies the dataset of Oxford-IIIT Pet Dataset (Parkhi et al. 2012). This dataset contains images of pets and their pixel-wise mask that indicates (1) pixels belonging to the pet, (2) pixels bordering the pet, and (3) surrounding pixels.

The Oxford-IIIT Pets dataset will be downloaded and unpacked through the cell blow.

if first_time_running:
    # downloading and executing data files
    import tarfile
    import urllib.request

    filename_image = filepath+'images.tar.gz'
    filename_target = filepath+'annotations.tar.gz'

    urllib.request.urlretrieve('http://www.robots.ox.ac.uk/~vgg/data/pets/data/images.tar.gz', filename_image);
    urllib.request.urlretrieve('https://www.robots.ox.ac.uk/~vgg/data/pets/data/annotations.tar.gz', filename_target);

    with tarfile.open(filename_image, "r:gz") as tar_io:
        tar_io.extractall(path=filepath)
    with tarfile.open(filename_target, "r:gz") as tar_io:
        tar_io.extractall(path=filepath)

    os.system(f'cd {filepath} && cd images && rm *.mat && cd../../')

import albumentations as albu

from swiss_army_keras.dataset_utils import SegmentationAlbumentationsDataLoader
from weathermentations.rain import DropsOnLens

n_labels = 3

input_tensor = tf.keras.layers.Input((256, 256, 3), batch_size=1)

X_decoder = models.deeplab_v3_plus(input_tensor,
                                   n_labels=n_labels,
                                   backbone='EfficientNetLiteB0',
                                   atrous_rates=[],
                                   deep_layer=5,
                                   activation='ReLU',
                                   batch_norm=True,
                                   pool=True,
                                   unpool=True,
                                   name='Pets')



OUT_stack = X_decoder(input_tensor)
deeplabv3plus = keras.models.Model([input_tensor,], OUT_stack)
deeplabv3plus.preprocessing = X_decoder.preprocessing
deeplabv3plus.summary()

KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'")
Model: "model_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
input_1 (InputLayer)         [(1, 256, 256, 3)]        0
_________________________________________________________________
model (Functional)           (1, 256, 256, 3)          4652771
=================================================================
Total params: 4,652,771
Trainable params: 2,021,731
Non-trainable params: 2,631,040
_________________________________________________________________

def get_training_augmentation(width=513, height=513):
    train_transform = [


        #albu.RandomRotate90(p=0.5),

        #albu.OneOf([
        #
        #
        #], p=1),
        #albu.Resize(width, height, 2, p=1),

        #albu.ShiftScaleRotate(shift_limit = 0.05, scale_limit=(0.1, 0.1), interpolation=2, rotate_limit=25, p=0.3),

        albu.HorizontalFlip(p=0.5),
        #albu.VerticalFlip(p=0.15),

        albu.Compose([
        albu.OneOf([
            albu.ISONoise(p=1),
            albu.MultiplicativeNoise(p=1),
            albu.GaussNoise(p=1),
            albu.ImageCompression(p=1),

            ],
            p=1,
        ),

        #albu.Perspective(pad_mode=2, p=0.1),

        albu.OneOf(
            [
                albu.CLAHE(p=1),
                albu.RandomBrightnessContrast(p=1),
                albu.Sharpen(p=1),
                albu.Emboss(p=1),
                albu.RandomGamma(p=1),
            ],
            p=0.5,
        ),

        albu.OneOf(
            [
                albu.Blur(blur_limit=5, p=1),
                albu.MedianBlur(blur_limit=5, p=1),
                albu.MotionBlur(blur_limit=5, p=1),
            ],
            p=0.25,
        ),

        albu.OneOf(
            [
                albu.RandomBrightnessContrast(p=1),
                albu.HueSaturationValue(p=1),
                albu.RGBShift(p=1),
            ],
            p=0.25,
        ),

        albu.OneOf(
            [
                #albu.ElasticTransform(alpha=10, sigma=10 * 0.05, alpha_affine=10 * 0.03, p=0.5),
                albu.GridDistortion(p=0.5),
                albu.OpticalDistortion(distort_limit=0.3, shift_limit=0.05, p=1)
            ],
            p=0.25,
        )], p=0.5),

        #DropsOnLens(maxR=40, minR=30, p=0.1),
        albu.ToGray(p=0.2)
    ]

    return albu.Compose(train_transform)
    #return albu.Compose([])


loader = SegmentationAlbumentationsDataLoader(num_classes=3,
                                              #precache=True,
                                              batch_size=16,
                                              width=256,
                                              height=256,
                                              dataset_path=filepath,
                                              masks_dir='annotations/trimaps',
                                              train_augmentations=get_training_augmentation(64, 64),
                                              label_shift=1,
                                              train_val_test_split=[0.8, 0.1, 0.1],
                                              normalization=deeplabv3plus.preprocessing,
                                              )

WARNING:root:Normalizing in range: [-1.  1.]

Train Images are good to go

WARNING:root:KEYS: None
WARNING:root:Router on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:KEYS: (b'P+S690P{iVPfx<aFJwxfSY^ugFzjuWOnaIh!o7J<', b':$80ST.hxA5xL7c+@$3YTEohOR^GhrJ2$qzN@bR^')
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Attaching HOST CURVE keys.
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 128, 'value': 93, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CLOSED'}
WARNING:root:KEYS: None
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4, 'value': 297, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECT_RETRIED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 128, 'value': 94, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CLOSED'}
WARNING:root:Router on "ipc:///tmp/swiss_army_keras_segmentation_curve_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4, 'value': 297, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECT_RETRIED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 8, 'value': 104, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_LISTENING'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_curve_socket_HruyKtc=): {'event': 8, 'value': 105, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_curve_socket_HruyKtc=', 'description': 'EVENT_LISTENING'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 95, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 106, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 96, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 107, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 98, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 108, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 98, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 111, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 99, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 109, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 93, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 101, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 122, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 94, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 110, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 101, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 124, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 102, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 112, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 113, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 103, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 127, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 104, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 114, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 115, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 106, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:KEYS: None
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 116, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 106, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 117, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 107, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 111, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 118, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 108, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 119, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 109, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:KEYS: None
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 134, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Dealer on "ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc="
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 32, 'value': 120, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_ACCEPTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 1, 'value': 110, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_CONNECTED'}
WARNING:root:Event Monitor(Host: ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=): {'event': 4096, 'value': 0, 'endpoint': b'ipc:///tmp/swiss_army_keras_segmentation_clear_socket_HruyKtc=', 'description': 'EVENT_HANDSHAKE_SUCCEEDED'}

train_set, val_set, test_set = loader.build_datasets()

loader.show_images(dset='train', num_images=16)

from swiss_army_keras import losses, metrics

loss = losses.loss_adapter(losses.symmetrical_unified_focal_loss(), from_logits=True, label_smoothing=0.1)
accuracy = tf.keras.metrics.CategoricalAccuracy()
miou = metrics.MeanIOU(num_classes=n_labels)

from swiss_army_keras.training_utils import TrainingDriver
from adabelief_tf import AdaBeliefOptimizer

td = TrainingDriver(deeplabv3plus,'pets_model', AdaBeliefOptimizer(), loss, [accuracy, miou], train_set, val_set, test_set, 20 )

Please check your arguments if you have upgraded adabelief-tf from version 0.0.1.
Modifications to default arguments:
                           eps  weight_decouple    rectify
-----------------------  -----  -----------------  -------------
adabelief-tf=0.0.1       1e-08  Not supported      Not supported
>=0.1.0 (Current 0.2.1)  1e-14  supported          default: True
SGD better than Adam (e.g. CNN for Image Classification)    Adam better than SGD (e.g. Transformer, GAN)
----------------------------------------------------------  ----------------------------------------------
Recommended epsilon = 1e-7                                  Recommended epsilon = 1e-14
For a complete table of recommended hyperparameters, see
https://github.com/juntang-zhuang/Adabelief-Optimizer
You can disable the log message by setting "print_change_log = False", though it is recommended to keep as a reminder.

td.run()

Model: "model_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
input_1 (InputLayer)         [(1, 256, 256, 3)]        0
_________________________________________________________________
model (Functional)           (1, 256, 256, 3)          4652771
=================================================================
Total params: 4,652,771
Trainable params: 2,021,731
Non-trainable params: 2,631,040
_________________________________________________________________
Epoch 1/20
WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

/usr/local/lib/python3.6/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
  category=CustomMaskWarning)
WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

369/369 [==============================] - ETA: 0s - loss: 0.4020 - categorical_accuracy: 0.8556 - mean_iou: 0.6432WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

369/369 [==============================] - 34s 72ms/step - loss: 0.4020 - categorical_accuracy: 0.8556 - mean_iou: 0.6432 - val_loss: 0.3942 - val_categorical_accuracy: 0.8722 - val_mean_iou: 0.6910
Epoch 2/20
369/369 [==============================] - 25s 68ms/step - loss: 0.3573 - categorical_accuracy: 0.8992 - mean_iou: 0.7252 - val_loss: 0.3513 - val_categorical_accuracy: 0.9059 - val_mean_iou: 0.7555
Epoch 3/20
369/369 [==============================] - 25s 68ms/step - loss: 0.3472 - categorical_accuracy: 0.9085 - mean_iou: 0.7450 - val_loss: 0.3609 - val_categorical_accuracy: 0.8975 - val_mean_iou: 0.7307
Epoch 4/20
369/369 [==============================] - 25s 68ms/step - loss: 0.3427 - categorical_accuracy: 0.9128 - mean_iou: 0.7541 - val_loss: 0.3413 - val_categorical_accuracy: 0.9147 - val_mean_iou: 0.7666
Epoch 5/20
369/369 [==============================] - 25s 68ms/step - loss: 0.3392 - categorical_accuracy: 0.9159 - mean_iou: 0.7611 - val_loss: 0.3381 - val_categorical_accuracy: 0.9161 - val_mean_iou: 0.7653
Epoch 6/20
369/369 [==============================] - 25s 68ms/step - loss: 0.3365 - categorical_accuracy: 0.9185 - mean_iou: 0.7671 - val_loss: 0.3359 - val_categorical_accuracy: 0.9190 - val_mean_iou: 0.7677
Epoch 7/20
369/369 [==============================] - 25s 68ms/step - loss: 0.3338 - categorical_accuracy: 0.9206 - mean_iou: 0.7719 - val_loss: 0.3448 - val_categorical_accuracy: 0.9122 - val_mean_iou: 0.7718
Epoch 8/20
369/369 [==============================] - 25s 68ms/step - loss: 0.3324 - categorical_accuracy: 0.9220 - mean_iou: 0.7749 - val_loss: 0.3363 - val_categorical_accuracy: 0.9185 - val_mean_iou: 0.7703
Epoch 9/20
369/369 [==============================] - 25s 69ms/step - loss: 0.3311 - categorical_accuracy: 0.9231 - mean_iou: 0.7774 - val_loss: 0.3293 - val_categorical_accuracy: 0.9246 - val_mean_iou: 0.7814
Epoch 10/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3290 - categorical_accuracy: 0.9247 - mean_iou: 0.7811 - val_loss: 0.3298 - val_categorical_accuracy: 0.9246 - val_mean_iou: 0.7860
Epoch 11/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3277 - categorical_accuracy: 0.9260 - mean_iou: 0.7841 - val_loss: 0.3350 - val_categorical_accuracy: 0.9196 - val_mean_iou: 0.7772
Epoch 12/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3270 - categorical_accuracy: 0.9264 - mean_iou: 0.7854 - val_loss: 0.3396 - val_categorical_accuracy: 0.9155 - val_mean_iou: 0.7709
Epoch 13/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3249 - categorical_accuracy: 0.9283 - mean_iou: 0.7895 - val_loss: 0.3311 - val_categorical_accuracy: 0.9233 - val_mean_iou: 0.7847
Epoch 14/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3248 - categorical_accuracy: 0.9283 - mean_iou: 0.7897 - val_loss: 0.3342 - val_categorical_accuracy: 0.9208 - val_mean_iou: 0.7768
Epoch 15/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3232 - categorical_accuracy: 0.9296 - mean_iou: 0.7924 - val_loss: 0.3303 - val_categorical_accuracy: 0.9235 - val_mean_iou: 0.7871
Epoch 16/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3226 - categorical_accuracy: 0.9302 - mean_iou: 0.7943 - val_loss: 0.3273 - val_categorical_accuracy: 0.9269 - val_mean_iou: 0.7877
Epoch 17/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3213 - categorical_accuracy: 0.9313 - mean_iou: 0.7963 - val_loss: 0.3287 - val_categorical_accuracy: 0.9251 - val_mean_iou: 0.7878
Epoch 18/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3207 - categorical_accuracy: 0.9319 - mean_iou: 0.7979 - val_loss: 0.3296 - val_categorical_accuracy: 0.9248 - val_mean_iou: 0.7900
Epoch 19/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3192 - categorical_accuracy: 0.9331 - mean_iou: 0.8008 - val_loss: 0.3265 - val_categorical_accuracy: 0.9270 - val_mean_iou: 0.7960
Epoch 20/20
369/369 [==============================] - 26s 70ms/step - loss: 0.3182 - categorical_accuracy: 0.9339 - mean_iou: 0.8023 - val_loss: 0.3279 - val_categorical_accuracy: 0.9259 - val_mean_iou: 0.7865

WARNING:root:Unfreezing Model

Model: "model_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
input_1 (InputLayer)         [(1, 256, 256, 3)]        0
_________________________________________________________________
model (Functional)           (1, 256, 256, 3)          4652771
=================================================================
Total params: 4,652,771
Trainable params: 4,611,299
Non-trainable params: 41,472
_________________________________________________________________
Epoch 1/20
WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

369/369 [==============================] - ETA: 0s - loss: 0.3725 - categorical_accuracy: 0.8885 - mean_iou: 0.7141WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

WARNING:tensorflow:Model was constructed with shape (1, 256, 256, 3) for input KerasTensor(type_spec=TensorSpec(shape=(1, 256, 256, 3), dtype=tf.float32, name='input_1'), name='input_1', description="created by layer 'input_1'"), but it was called on an input with incompatible shape (16, 256, 256, 3).

369/369 [==============================] - 79s 158ms/step - loss: 0.3725 - categorical_accuracy: 0.8885 - mean_iou: 0.7141 - val_loss: 0.3536 - val_categorical_accuracy: 0.9022 - val_mean_iou: 0.7398
Epoch 2/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3466 - categorical_accuracy: 0.9086 - mean_iou: 0.7514 - val_loss: 0.3413 - val_categorical_accuracy: 0.9133 - val_mean_iou: 0.7657
Epoch 3/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3382 - categorical_accuracy: 0.9162 - mean_iou: 0.7665 - val_loss: 0.3346 - val_categorical_accuracy: 0.9198 - val_mean_iou: 0.7775
Epoch 4/20
369/369 [==============================] - 55s 148ms/step - loss: 0.3327 - categorical_accuracy: 0.9211 - mean_iou: 0.7763 - val_loss: 0.3299 - val_categorical_accuracy: 0.9233 - val_mean_iou: 0.7848
Epoch 5/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3298 - categorical_accuracy: 0.9237 - mean_iou: 0.7820 - val_loss: 0.3296 - val_categorical_accuracy: 0.9239 - val_mean_iou: 0.7882
Epoch 6/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3274 - categorical_accuracy: 0.9256 - mean_iou: 0.7861 - val_loss: 0.3270 - val_categorical_accuracy: 0.9259 - val_mean_iou: 0.7867
Epoch 7/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3244 - categorical_accuracy: 0.9285 - mean_iou: 0.7923 - val_loss: 0.3286 - val_categorical_accuracy: 0.9250 - val_mean_iou: 0.7894
Epoch 8/20
369/369 [==============================] - 54s 146ms/step - loss: 0.3231 - categorical_accuracy: 0.9295 - mean_iou: 0.7942 - val_loss: 0.3310 - val_categorical_accuracy: 0.9223 - val_mean_iou: 0.7835
Epoch 9/20
369/369 [==============================] - 55s 148ms/step - loss: 0.3209 - categorical_accuracy: 0.9314 - mean_iou: 0.7983 - val_loss: 0.3263 - val_categorical_accuracy: 0.9270 - val_mean_iou: 0.7944
Epoch 10/20
369/369 [==============================] - 54s 145ms/step - loss: 0.3197 - categorical_accuracy: 0.9324 - mean_iou: 0.8006 - val_loss: 0.3268 - val_categorical_accuracy: 0.9261 - val_mean_iou: 0.7907
Epoch 11/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3182 - categorical_accuracy: 0.9337 - mean_iou: 0.8034 - val_loss: 0.3222 - val_categorical_accuracy: 0.9304 - val_mean_iou: 0.7989
Epoch 12/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3183 - categorical_accuracy: 0.9338 - mean_iou: 0.8039 - val_loss: 0.3264 - val_categorical_accuracy: 0.9273 - val_mean_iou: 0.7977
Epoch 13/20
369/369 [==============================] - 54s 148ms/step - loss: 0.3168 - categorical_accuracy: 0.9350 - mean_iou: 0.8064 - val_loss: 0.3264 - val_categorical_accuracy: 0.9276 - val_mean_iou: 0.7952
Epoch 14/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3149 - categorical_accuracy: 0.9365 - mean_iou: 0.8101 - val_loss: 0.3242 - val_categorical_accuracy: 0.9292 - val_mean_iou: 0.7985
Epoch 15/20
369/369 [==============================] - 55s 148ms/step - loss: 0.3142 - categorical_accuracy: 0.9371 - mean_iou: 0.8115 - val_loss: 0.3274 - val_categorical_accuracy: 0.9272 - val_mean_iou: 0.7959
Epoch 16/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3145 - categorical_accuracy: 0.9367 - mean_iou: 0.8109 - val_loss: 0.3215 - val_categorical_accuracy: 0.9309 - val_mean_iou: 0.8016
Epoch 17/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3130 - categorical_accuracy: 0.9380 - mean_iou: 0.8135 - val_loss: 0.3233 - val_categorical_accuracy: 0.9295 - val_mean_iou: 0.7984
Epoch 18/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3117 - categorical_accuracy: 0.9391 - mean_iou: 0.8162 - val_loss: 0.3201 - val_categorical_accuracy: 0.9322 - val_mean_iou: 0.8038
Epoch 19/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3136 - categorical_accuracy: 0.9375 - mean_iou: 0.8128 - val_loss: 0.3252 - val_categorical_accuracy: 0.9280 - val_mean_iou: 0.7962
Epoch 20/20
369/369 [==============================] - 54s 147ms/step - loss: 0.3119 - categorical_accuracy: 0.9389 - mean_iou: 0.8158 - val_loss: 0.3201 - val_categorical_accuracy: 0.9322 - val_mean_iou: 0.8035

WARNING:root:Quantizing Model

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

INFO:tensorflow:Assets written to: /tmp/tmpuq2_6gro/assets

INFO:tensorflow:Assets written to: /tmp/tmpuq2_6gro/assets
/usr/local/lib/python3.6/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
  category=CustomMaskWarning)
WARNING:absl:For model inputs containing unsupported operations which cannot be quantized, the `inference_input_type` attribute will default to the original type.

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.

INFO:tensorflow:Assets written to: /tmp/tmpobro38yf/assets

INFO:tensorflow:Assets written to: /tmp/tmpobro38yf/assets
/usr/local/lib/python3.6/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
  category=CustomMaskWarning)

<Figure size 432x288 with 0 Axes>

loader.show_results(deeplabv3plus, 16)

Model User Guide

This user guide requires `keras-unet-collection==0.1.9` or higher.

## Content

* [**U-net**](#U-net)
* [**V-net**](#V-net)
* [**Attention-Unet**](#Attention-Unet)
* [**U-net++**](#U-net++)
* [**UNET 3+**](#UNET-3+)
* [**R2U-net**](#R2U-net)
* [**ResUnet-a**](#ResUnet-a)
* [**U^2-Net**](#U^2-Net)
* [**TransUNET**](#TransUNET)
* [**Swin-UNET**](#Swin-UNET)

import tensorflow as tf
from tensorflow import keras

print('TensorFlow {}; Keras {}'.format(tf.__version__, keras.__version__))

TensorFlow 2.5.0; Keras 2.5.0

Step 1: importing models from swiss_army_keras

from swiss_army_keras import models

Step 2: defining your hyper-parameters

Commonly used hyper-parameter options are listed as follows. Full details are available through the Python helper function:

• inpust_size: a tuple or list that defines the shape of input tensors.

  • models.resunet_a_2d, models.transunet_2d, and models.swin_unet_2d support int only, others also support inpust_size=(None, None, 3).

  • activation='PReLU' is not compatible with inpust_size=(None, None, 3).

• filter_num: a list that defines the number of convolutional filters per down- and up-sampling blocks.

  • For unet_2d, att_unet_2d, unet_plus_2d, r2_unet_2d, depth \(\ge\) 2 is expected.

  • For resunet_a_2d and u2net_2d, depth \(\ge\) 3 is expected.

• n_labels: number of output targets, e.g., n_labels=2 for binary classification.

• activation: the activation function of hidden layers. Available choices are 'ReLU', 'LeakyReLU', 'PReLU', 'ELU', 'GELU', 'Snake'.

• output_activation: the activation function of the output layer. Recommended choices are 'Sigmoid', 'Softmax', None (linear), 'Snake'.

• batch_norm: if specified as True, all convolutional layers will be configured as stacks of “Conv2D-BN-Activation”.

• stack_num_down: number of convolutional layers per downsampling level.

• stack_num_up: number of convolutional layers (after concatenation) per upsampling level.

• pool: the configuration of downsampling (encoding) blocks.

  • pool=False: downsampling with a convolutional layer (2-by-2 convolution kernels with 2 strides; optional batch normalization and activation).

  • pool=True or pool='max' downsampling with a max-pooling layer.

  • pool='ave' downsampling with a average-pooling layer.

• unpool: the configuration of upsampling (decoding) blocks.

  • unpool=False: upsampling with a transpose convolutional layer (2-by-2 convolution kernels with 2 strides; optional batch normalization and activation).

  • unpool=True or unpool='bilinear' upsampling with bilinear interpolation.

  • unpool='nearest' upsampling with reflective padding.

• name: user-specified prefix of the configured layer and model. Use keras.models.Model.summary to identify the exact name of each layer.

Step 3: Configuring your model

Note

Configured models can be saved through model.save(filepath, save_traces=True), but they may contain python objects that are not part of the tensorflow.keras. Thus when loading the model, it is preferred to load the weights only, and set/freeze them within a new configuration.

e.g.

weights = dummy_loader(model_old_path)
model_new = swin_transformer_model(...)
model_new.set_weights(weights)

U-net

Example 1: U-net for binary classification with:

1. Five down- and upsampliung levels (or four downsampling levels and one bottom level).

2. Two convolutional layers per downsampling level.

3. One convolutional layer (after concatenation) per upsamling level.

4. Gaussian Error Linear Unit (GELU) activcation, Softmax output activation, batch normalization.

5. Downsampling through Maxpooling.

6. Upsampling through reflective padding.

model = models.unet_2d((None, None, 3), [64, 128, 256, 512, 1024], n_labels=2,
                      stack_num_down=2, stack_num_up=1,
                      activation='GELU', output_activation='Softmax',
                      batch_norm=True, pool='max', unpool='nearest', name='unet')

V-net

Example 2: Vnet (originally proposed for 3-d inputs, here modified for 2-d inputs) for binary classification with:

1. Input size of (256, 256, 1); PReLU does not support input tensor with shapes of NoneType

2. Five down- and upsampliung levels (or four downsampling levels and one bottom level).

3. Number of stacked convolutional layers of the residual path increase with downsampling levels from one to three (symmetrically, decrease with upsampling levels).

   • res_num_ini=1

   • res_num_max=3

4. PReLU activcation, Softmax output activation, batch normalization.

5. Downsampling through stride convolutional layers.

6. Upsampling through transpose convolutional layers.

model = models.vnet_2d((256, 256, 1), filter_num=[16, 32, 64, 128, 256], n_labels=2,
                      res_num_ini=1, res_num_max=3,
                      activation='PReLU', output_activation='Softmax',
                      batch_norm=True, pool=False, unpool=False, name='vnet')

Attention-Unet

Example 3: attention-Unet for single target regression with:

1. Four down- and upsampling levels.

2. Two convolutional layers per downsampling level.

3. Two convolutional layers (after concatenation) per upsampling level.

4. ReLU activation, linear output activation (None), batch normalization.

5. Additive attention, ReLU attention activation.

6. Downsampling through stride convolutional layers.

7. Upsampling through bilinear interpolation.

model = models.att_unet_2d((None, None, 3), [64, 128, 256, 512], n_labels=1,
                           stack_num_down=2, stack_num_up=2,
                           activation='ReLU', atten_activation='ReLU', attention='add', output_activation=None,
                           batch_norm=True, pool=False, unpool='bilinear', name='attunet')

U-net++

Example 4: U-net++ for three-label classification with:

1. Four down- and upsampling levels.

2. Two convolutional layers per downsampling level.

3. Two convolutional layers (after concatenation) per upsampling level.

4. LeakyReLU activation, Softmax output activation, no batch normalization.

5. Downsampling through Maxpooling.

6. Upsampling through transpose convolutional layers.

7. Deep supervision.

model = models.unet_plus_2d((None, None, 3), [64, 128, 256, 512], n_labels=3,
                            stack_num_down=2, stack_num_up=2,
                            activation='LeakyReLU', output_activation='Softmax',
                            batch_norm=False, pool='max', unpool=False, deep_supervision=True, name='xnet')

----------
deep_supervision = True
names of output tensors are listed as follows ("sup0" is the shallowest supervision layer;
"final" is the final output layer):

        xnet_output_sup0_activation
        xnet_output_sup1_activation
        xnet_output_sup2_activation
        xnet_output_final_activation

UNET 3+

Example 5: UNet 3+ for binary classification with:

1. Four down- and upsampling levels.

2. Two convolutional layers per downsampling level.

3. One convolutional layers (after concatenation) per upsampling level.

4. ReLU activation, Sigmoid output activation, batch normalization.

5. Downsampling through Maxpooling.

6. Upsampling through transpose convolutional layers.

7. Deep supervision.

model = models.unet_3plus_2d((128, 128, 3), n_labels=2, filter_num_down=[64, 128, 256, 512],
                             filter_num_skip='auto', filter_num_aggregate='auto',
                             stack_num_down=2, stack_num_up=1, activation='ReLU', output_activation='Sigmoid',
                             batch_norm=True, pool='max', unpool=False, deep_supervision=True, name='unet3plus')

Automated hyper-parameter determination is applied with the following details:
----------
        Number of convolution filters after each full-scale skip connection: filter_num_skip = [64, 64, 64]
        Number of channels of full-scale aggregated feature maps: filter_num_aggregate = 256
----------
deep_supervision = True
names of output tensors are listed as follows ("sup0" is the shallowest supervision layer;
"final" is the final output layer):

        unet3plus_output_sup0_activation
        unet3plus_output_sup1_activation
        unet3plus_output_sup2_activation
        unet3plus_output_final_activation

• filter_num_skip and filter_num_aggregate can be specified explicitly:

model = models.unet_3plus_2d((128, 128, 3), n_labels=2, filter_num_down=[64, 128, 256, 512],
                             filter_num_skip=[64, 64, 64], filter_num_aggregate=256,
                             stack_num_down=2, stack_num_up=1, activation='ReLU', output_activation='Sigmoid',
                             batch_norm=True, pool='max', unpool=False, deep_supervision=True, name='unet3plus')

----------
deep_supervision = True
names of output tensors are listed as follows ("sup0" is the shallowest supervision layer;
"final" is the final output layer):

        unet3plus_output_sup0_activation
        unet3plus_output_sup1_activation
        unet3plus_output_sup2_activation
        unet3plus_output_final_activation

R2U-net

Example 6: R2U-net for binary classification with:

1. Four down- and upsampling levels.

2. Two recurrent convolutional layers with two iterations per down- and upsampling level.

3. ReLU activation, Softmax output activation, no batch normalization.

4. Downsampling through Maxpooling.

5. Upsampling through reflective padding.

model = models.r2_unet_2d((None, None, 3), [64, 128, 256, 512], n_labels=2,
                          stack_num_down=2, stack_num_up=1, recur_num=2,
                          activation='ReLU', output_activation='Softmax',
                          batch_norm=True, pool='max', unpool='nearest', name='r2unet')

ResUnet-a

Example 7: ResUnet-a for 16-label classification with:

1. input size of (128, 128, 3)

2. Six downsampling levels followed by an Atrous Spatial Pyramid Pooling (ASPP) layer with 256 filters.

3. Six upsampling levels followed by an ASPP layer with 128 filters.

4. dilation rates of {1, 3, 15, 31} for shallow layers, {1,3,15} for intermediate layers, and {1,} for deep layers.

5. ReLU activation, Sigmoid output activation, batch normalization.

6. Downsampling through stride convolutional layers.

7. Upsampling through reflective padding.

model = models.resunet_a_2d((128, 128, 3), [32, 64, 128, 256, 512, 1024],
                            dilation_num=[1, 3, 15, 31],
                            n_labels=16, aspp_num_down=256, aspp_num_up=128,
                            activation='ReLU', output_activation='Sigmoid',
                            batch_norm=True, pool=False, unpool='nearest', name='resunet')

Received dilation rates: [1, 3, 15, 31]
Received dilation rates are not defined on a per downsampling level basis.
Automated determinations are applied with the following details:
        depth-0, dilation_rate = [1, 3, 15, 31]
        depth-1, dilation_rate = [1, 3, 15, 31]
        depth-2, dilation_rate = [1, 3, 15]
        depth-3, dilation_rate = [1, 3, 15]
        depth-4, dilation_rate = [1]
        depth-5, dilation_rate = [1]

• dilation_num can be specified per down- and uplampling level:

model = models.resunet_a_2d((128, 128, 3), [32, 64, 128, 256, 512, 1024],
                            dilation_num=[[1, 3, 15, 31], [1, 3, 15, 31], [1, 3, 15], [1, 3, 15], [1,], [1,],],
                            n_labels=16, aspp_num_down=256, aspp_num_up=128,
                            activation='ReLU', output_activation='Sigmoid',
                            batch_norm=True, pool=False, unpool='nearest', name='resunet')

U^2-Net

Example 8: U^2-Net for binary classification with:

1. Six downsampling levels with the first four layers built with RSU, and the last two (one downsampling layer, one bottom layer) built with RSU-F4.

   • filter_num_down=[64, 128, 256, 512]

   • filter_mid_num_down=[32, 32, 64, 128]

   • filter_4f_num=[512, 512]

   • filter_4f_mid_num=[256, 256]

2. Six upsampling levels with the deepest layer built with RSU-F4, and the other four layers built with RSU.

   • filter_num_up=[64, 64, 128, 256]

   • filter_mid_num_up=[16, 32, 64, 128]

3. ReLU activation, Sigmoid output activation, batch normalization.

4. Deep supervision

5. Downsampling through stride convolutional layers.

6. Upsampling through transpose convolutional layers.

*In the original work of U^2-Net, down- and upsampling were achieved through maxpooling (pool=True or pool='max') and bilinear interpolation (unpool=True or unpool='bilinear').

model = models.u2net_2d((128, 128, 3), n_labels=2,
                        filter_num_down=[64, 128, 256, 512], filter_num_up=[64, 64, 128, 256],
                        filter_mid_num_down=[32, 32, 64, 128], filter_mid_num_up=[16, 32, 64, 128],
                        filter_4f_num=[512, 512], filter_4f_mid_num=[256, 256],
                        activation='ReLU', output_activation=None,
                        batch_norm=True, pool=False, unpool=False, deep_supervision=True, name='u2net')

----------
The depth of u2net_2d = len(filter_num_down) + len(filter_4f_num) = 6
----------
deep_supervision = True
names of output tensors are listed as follows ("sup0" is the shallowest supervision layer;
"final" is the final output layer):

        u2net_output_sup0_trans_conv
        u2net_output_sup1_trans_conv
        u2net_output_sup2_trans_conv
        u2net_output_sup3_trans_conv
        u2net_output_sup4_trans_conv
        u2net_output_sup5_trans_conv
        u2net_output_final

• u2net_2d supports automated determination of filter numbers per down- and upsampling level. Auto-mode may produce a slightly larger network.

model = models.u2net_2d((None, None, 3), n_labels=2,
                        filter_num_down=[64, 128, 256, 512],
                        activation='ReLU', output_activation='Sigmoid',
                        batch_norm=True, pool=False, unpool=False, deep_supervision=True, name='u2net')

Automated hyper-parameter determination is applied with the following details:
----------
        Number of RSU output channels within downsampling blocks: filter_num_down = [64, 128, 256, 512]
        Number of RSU intermediate channels within downsampling blocks: filter_mid_num_down = [16, 32, 64, 128]
        Number of RSU output channels within upsampling blocks: filter_num_up = [64, 128, 256, 512]
        Number of RSU intermediate channels within upsampling blocks: filter_mid_num_up = [16, 32, 64, 128]
        Number of RSU-4F output channels within downsampling and bottom blocks: filter_4f_num = [512, 512]
        Number of RSU-4F intermediate channels within downsampling and bottom blocks: filter_4f_num = [256, 256]
----------
Explicitly specifying keywords listed above if their "auto" settings do not satisfy your needs
----------
The depth of u2net_2d = len(filter_num_down) + len(filter_4f_num) = 6
----------
deep_supervision = True
names of output tensors are listed as follows ("sup0" is the shallowest supervision layer;
"final" is the final output layer):

        u2net_output_sup0_activation
        u2net_output_sup1_activation
        u2net_output_sup2_activation
        u2net_output_sup3_activation
        u2net_output_sup4_activation
        u2net_output_sup5_activation
        u2net_output_final_activation

TransUNET

Example 9: TransUNET for 12-label classification with:

• input size of (512, 512, 3)

• Four down- and upsampling levels.

• Two convolutional layers per downsampling level.

• Two convolutional layers (after concatenation) per upsampling level.

• 12 transformer blocks (num_transformer=12).

• 12 attention heads (num_heads=12).

• 3072 MLP nodes per vision transformer (num_mlp=3072).

• 768 embeding dimensions (embed_dim=768).

• Gaussian Error Linear Unit (GELU) activcation for transformer MLPs.

• ReLU activation, softmax output activation, batch normalization.

• Downsampling through maxpooling.

• Upsampling through bilinear interpolation.

model = models.transunet_2d((512, 512, 3), filter_num=[64, 128, 256, 512], n_labels=12, stack_num_down=2, stack_num_up=2,
                                embed_dim=768, num_mlp=3072, num_heads=12, num_transformer=12,
                                activation='ReLU', mlp_activation='GELU', output_activation='Softmax',
                                batch_norm=True, pool=True, unpool='bilinear', name='transunet')

Swin-UNET

Example 10: Swin-UNET for 3-label classification with:

• input size of (128, 128, 3)

• Four down- and upsampling levels (or three downsampling levels and one bottom level) (depth=4).

• Two Swin-Transformers per downsampling level.

• Two Swin-Transformers (after concatenation) per upsampling level.

• Extract 2-by-2 patches from the input (patch_size=(2, 2))

• Embed 2-by-2 patches to 64 dimensions (filter_num_begin=64, a.k.a, number of embedded dimensions).

• Number of attention heads for each down- and upsampling level: num_heads=[4, 8, 8, 8].

• Size of attention windows for each down- and upsampling level: window_size=[4, 2, 2, 2].

• 512 nodes per Swin-Transformer (num_mlp=512)

• Shift attention windows (i.e., Swin-MSA) (shift_window=True).

model = models.swin_unet_2d((128, 128, 3), filter_num_begin=64, n_labels=3, depth=4, stack_num_down=2, stack_num_up=2,
                            patch_size=(2, 2), num_heads=[4, 8, 8, 8], window_size=[4, 2, 2, 2], num_mlp=512,
                            output_activation='Softmax', shift_window=True, name='swin_unet')

Indices and tables

• Index

• Module Index

• Search Page