trainmixed.py

import numpy as np
import tensorflow as tf
from tensorflow import keras
import matplotlib.pyplot as plt
import PIL
import os


from tensorflow.keras import layers
from tensorflow.keras.models import Sequential

# Read and split data
batch_size = 16
img_height = 519
img_width = 695

train_dir_low = 'labeled_splits_low/train'
val_dir_low = 'labeled_splits_low/val'
train_dir_blo = 'labeled_splits_blo/train'
val_dir_blo = 'labeled_splits_blo/val'

train_ds_low = keras.utils.image_dataset_from_directory(
    train_dir_low, 
    batch_size=batch_size, 
    image_size=(img_height, img_width)
)

val_ds_low = keras.utils.image_dataset_from_directory(
    val_dir_low, 
    batch_size=batch_size, 
    image_size=(img_height, img_width)
)


train_ds_blo = keras.utils.image_dataset_from_directory(
    train_dir_blo, 
    batch_size=batch_size, 
    image_size=(img_height, img_width)
)

val_ds_blo = keras.utils.image_dataset_from_directory(
    val_dir_blo, 
    batch_size=batch_size, 
    image_size=(img_height, img_width)
)


train_ds = train_ds_low.concatenate(train_ds_blo)
val_ds = val_ds_low.concatenate(val_ds_blo)

train_ds.class_names = train_ds_low.class_names
val_ds.class_names = val_ds_low.class_names
class_names = train_ds.class_names
num_classes = len(class_names)


# Augment data
data_augmentation = keras.Sequential(
    [
        layers.RandomFlip("horizontal",
                            input_shape=(img_height,
                                        img_width,
                                        3)),
        layers.RandomRotation(0.1)
    ]
)

# Specify model
model = Sequential([
    data_augmentation,
    layers.Rescaling(1./255, input_shape=(img_height, img_width, 3)),
    layers.CenterCrop(280, 280),
    layers.Conv2D(16, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(32, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(64, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Flatten(),
    layers.Dropout(0.1),
    layers.Dense(128, activation='relu'),
    layers.Dense(num_classes)
])

# Compile model
model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])
model.summary()


checkpoint_path = "training_checkpoints/cp-{epoch:04d}"

# Callbacks
mc = tf.keras.callbacks.ModelCheckpoint(
    filepath=checkpoint_path, 
    verbose=1,
    save_freq='epoch')
cl = tf.keras.callbacks.CSVLogger('history.csv')
es = tf.keras.callbacks.EarlyStopping(patience=7, start_from_epoch=10)

# Load checkpoint if want to continue from checkpoints
continue_epoch = 37
model = tf.keras.models.load_model(f"training_checkpoints/cp-{continue_epoch:04d}")

epochs=50
history = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=epochs,
    initial_epoch=continue_epoch,
    batch_size=1,
    callbacks=[cl, mc, es],
    verbose=1
)