keras_mb_works

# import statements
import tensorflow as tf
from keras.preprocessing.image import ImageDataGenerator
import numpy as np


# blue empty white yellow

# loading training data
train_datagen = ImageDataGenerator(
        rescale=1./255,
        shear_range=0.1,
        zoom_range=0.1,
        brightness_range=[0.2, 1.0],
        horizontal_flip=True)
train_generator = train_datagen.flow_from_directory(
        'D:\\Scripts\\OpenCV\\ShelfData\\TrainWithSub\\',
        target_size=(30, 30),
        batch_size=56,
        class_mode='categorical')


cnn = tf.keras.models.Sequential()
cnn.add(tf.keras.layers.Conv2D(filters=48, kernel_size=3, activation='relu', input_shape=[30, 30, 3]))  # out shape (None, 28, 28, 48)
cnn.add(tf.keras.layers.MaxPool2D(pool_size=2, strides=2))  # out shape (None, 14, 14, 48)
cnn.add(tf.keras.layers.Conv2D(filters=48, kernel_size=3, activation='relu'))  # out shape (None, 12, 12, 48)
cnn.add(tf.keras.layers.MaxPool2D(pool_size=2, strides=2))  # out shape (None, 6, 6, 48)
cnn.add(tf.keras.layers.Conv2D(filters=32, kernel_size=3, activation='relu'))  # out shape (None, 4, 4, 32)
cnn.add(tf.keras.layers.MaxPool2D(pool_size=2, strides=2))  # out shape (None, 2, 2, 32)
cnn.add(tf.keras.layers.Flatten())
cnn.add(tf.keras.layers.Dense(128, activation='relu'))
cnn.add(tf.keras.layers.Dense(64, activation='relu'))
cnn.add(tf.keras.layers.Dense(4, activation='softmax'))

cnn.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])
cnn.fit(x=train_generator, epochs=14)

cnn.save('D:\\Scripts\\OpenCV\\ShelfData\\model_saved_5.h5')