6_DeepLabv3+_Road_CE_Loss.py

import numpy as np
import matplotlib.pyplot as plt
import os 
import cv2
import glob as glob
import albumentations as A 
import requests
import zipfile
import time
import pathlib

import tensorflow as tf
from tensorflow.keras.models import Model
from tensorflow.keras.utils import Sequence
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, AveragePooling2D, Conv2DTranspose, BatchNormalization, Activation, Dropout, Upsampling2D, Concatenate
from tensorflow.keras.applications.resnet50 import ResNet50, preprocess_input
from matplotlib.ticker
import MultipleLocator, FormatStrFormatter
from dataclasses import dataclass

def system_config():

@dataclass(frozen = True)
class DatasetConfig:

@dataclass(frozen = True)
class TrainingConfig:

@dataclass(frozen = True)
class InferenceConfig:

def convolution_block():

def DilatedSpatialPyramidPooling():

def deeplabv3plus():

model = deeplabv3plus()
model.summary()

def download_file()

def unzip()

save_name

class CustomSegDataLoader():
    def __init__():

    def __len__():

    def transforms():

    def resize():

    def reset_array():

    def __getitem__():


id2color = {}
id2color_display = {}

def rgb_to_onehot():

def num_to_rgb():

def image_overlay():

def display_image_and_mask():

def create_datasets():

train_ds, valid_ds = 

for i, (images,masks) in enumerate(train_ds):
    if i ==3:
        break
    image, mask = images[0], masks[0]
    display_image_and_mask()

def mean_iou():
    return mean_iou

model.compile()

if not os.path.exists(TrainingConfig.CHECKPOINT_DIR):
    os.makedirs()
num_versions
version_dir
os.makedirs
model_checkpoint_callback

history = model.fit()

def plot_results():
    plt.close()

train_acc
valid_acc

train_iou
valid_iou

plot_results
plot_results

train_loss
valid_loss

max_loss

plot_results

trained_model
evaluate 
print
print

def inference():
inference()