emotions.py

# -*- coding: utf-8 -*-
"""Emotions.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1PWL-JvW1u7GH4SsvOjHfwq55vrDNpXt8
"""

import tensorflow as tf
from keras.preprocessing.image import ImageDataGenerator

# Commented out IPython magic to ensure Python compatibility.
# %%bash
# pwd
# cd drive/'My Drive'/'Colab Notebooks'/'Deep Learning'/'Convolutional neural network'/'facial expressions'
# cp emodata.zip /content

# Commented out IPython magic to ensure Python compatibility.
# %%bash
# sudo apt-get install unzip
# unzip emodata.zip

train_datagen = ImageDataGenerator(
        rescale=1./255,  #(feature-scaling)
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True)

training_set = train_datagen.flow_from_directory('/content/train',
        target_size=(128, 128),
        batch_size=32,
        color_mode='grayscale',
        class_mode='categorical',
        shuffle=True)

test_datagen = ImageDataGenerator(rescale=1./255)

testing_set = test_datagen.flow_from_directory('/content/validation',
        target_size=(128, 128),
        batch_size=32,
        color_mode='grayscale',
        class_mode='categorical')

cnn = tf.keras.models.Sequential()

cnn.add(tf.keras.layers.Conv2D(filters=32, kernel_size=3, activation='relu', input_shape=[128, 128, 1]))
cnn.add(tf.keras.layers.MaxPool2D(pool_size=2, strides=2))
cnn.add(tf.keras.layers.Dropout(0.2))

cnn.add(tf.keras.layers.Conv2D(filters=32, kernel_size=3, activation='relu'))
cnn.add(tf.keras.layers.MaxPool2D(pool_size=2, strides=2))
cnn.add(tf.keras.layers.Dropout(0.2))

cnn.add(tf.keras.layers.Conv2D(filters=32, kernel_size=3, activation='relu'))
cnn.add(tf.keras.layers.MaxPool2D(pool_size=2, strides=2))
cnn.add(tf.keras.layers.Dropout(0.2))

cnn.add(tf.keras.layers.Conv2D(filters=32, kernel_size=3, activation='relu'))
cnn.add(tf.keras.layers.MaxPool2D(pool_size=2, strides=2))
cnn.add(tf.keras.layers.Dropout(0.2))

cnn.add(tf.keras.layers.Flatten())

cnn.add(tf.keras.layers.Dense(units=128, activation='relu'))

cnn.add(tf.keras.layers.Dense(units=7, activation='softmax'))

cnn.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

train_samples = 28821
test_samples = 7066

cnn.fit_generator(training_set, steps_per_epoch=train_samples//32, epochs=25, validation_data=testing_set, validation_steps=test_samples//32 )

cnn.save('emotions_5layer.h5')