CNN section added

Author: Nikronic

Part 8 - Deep Learning/Section 22 - Convolutional Neural Networks (CNN)/cnn.py

@@ -0,0 +1,74 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Mon Sep 24 22:48:03 2018
+
+@author: Mohammad Doosti Lakhani
+"""
+
+
+"""Important Note: This implementation take about hours on CPU. Use GPU or colab.research.google.com"""
+
+from keras.models import Sequential
+from keras.layers import Conv2D
+from keras.layers import MaxPool2D
+from keras.layers import Dense
+from keras.layers import Flatten
+
+
+# Preprocessing
+"""In this step, we just need to put training and test files to the folders with below structure :
+    -test_set
+        ->class1
+        ->class2
+        ->...
+    -training_set
+        ->class1
+        ->class2
+        ->...
+    """
+
+    
+# Defining some control parameters    
+image_x = 128
+image_y = 128
+image_channels = 3
+
+training_set_path = 'dataset/training_set'
+test_set_path = 'dataset/test_set'
+    
+model = Sequential() # Building sequential model
+
+# First layer of convolution
+model.add(Conv2D(64, kernel_size=(5,5), strides=(1,1), activation='relu', input_shape=(image_x,image_y,image_channels)))
+model.add(MaxPool2D(strides=(2,2), pool_size=(2,2)))
+
+# Second layer of convolution
+model.add(Conv2D(64, kernel_size=(5,5), strides=(1,1), activation='relu'))
+model.add(MaxPool2D(strides=(2,2), pool_size=(2,2)))
+
+# Flatten convolved tensors for feeding as input to Fully Connected layer
+model.add(Flatten())
+
+# First Hidden layer of fully connected
+model.add(Dense(units=256, activation='relu'))
+
+# Second Hidden layer of fully connected
+model.add(Dense(units=1, activation='sigmoid'))
+
+# Compiling the model
+model.compile(optimizer='adam', metrics=['accuracy'], loss='binary_crossentropy')
+
+# Data Augmentation
+# Because we have small dataset, to prevent overfitting and train better, we use this method
+# We do augmentation on both test_set and training_set
+
+from keras.preprocessing.image import ImageDataGenerator
+
+train_datagen = ImageDataGenerator(rescale = 1./255, shear_range = 0.2, zoom_range = 0.2, horizontal_flip = True)
+test_datagen = ImageDataGenerator(rescale = 1./255)
+training_set = train_datagen.flow_from_directory(training_set_path, target_size = (image_x, image_y), batch_size = 32, class_mode = 'binary')
+test_set = test_datagen.flow_from_directory(test_set_path, target_size = (image_x, image_y), batch_size = 32, class_mode = 'binary')
+
+# Firring model with original and augmented data
+model.fit_generator(training_set, steps_per_epoch = 8000, epochs = 25, validation_data = test_set, validation_steps = 2000)
+