Created
August 28, 2019 13:49
-
-
Save zckkte/b2465e36295f024356730e0d67b034d6 to your computer and use it in GitHub Desktop.
Cat vs dog image classifier
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import os | |
import math | |
import random | |
import shutil | |
import numpy as np | |
import matplotlib.pyplot as plt | |
from time import time | |
from functional import seq | |
from collections import namedtuple | |
import keras | |
from keras import layers | |
from keras import optimizers | |
from keras.regularizers import l1 | |
from keras.preprocessing.image import ImageDataGenerator | |
EPOCHS = 80 | |
BATCH_SIZE = 64 | |
IMG_HEIGHT, IMG_WIDTH=(150, 150) | |
ClassSplit = namedtuple('ClassSplit', 'class_label name split_label') | |
CopyVector = namedtuple('CopyVector', 'src dest') | |
def partition_on_ratio(entries, ratio): | |
test_ratio, val_ratio, train_ratio = ratio | |
split_train_index = int(train_ratio * len(entries)) | |
split_val_index = split_train_index + int(val_ratio * len(entries)) | |
files_train = entries[:split_train_index] | |
files_val = entries[split_train_index:split_val_index] | |
files_test = entries[split_val_index:] | |
return ('test', files_test), ('validation', files_val), ('training', files_train) | |
def class_split_to_copy_vec(base, class_split): | |
src = os.path.join(base, '{}'.format(class_split.name)) | |
dest = os.path.join(base, '{}/{}/{}'.format(class_split.split_label, class_split.class_label, class_split.name)) | |
return CopyVector(src=src, dest=dest) | |
def copy_makedir(copy_vector): | |
dest_dir = os.path.dirname(copy_vector.dest) | |
if not os.path.exists(dest_dir): | |
os.makedirs(dest_dir) | |
print(copy_vector.src) | |
print(copy_vector.dest) | |
shutil.copyfile(copy_vector.src, copy_vector.dest) | |
def split_files(src='dataset', ratio=(0.15, 0.15, 0.70)): | |
base = os.path.join(os.getcwd(), src) | |
test, validation, training = partition_on_ratio(os.listdir(base), ratio) | |
get_label = lambda x : 'cats' if 'cat' in x else 'dogs' | |
copy_vectors = seq([test, validation, training]) \ | |
.flat_map(lambda r : seq(r[1]).map(lambda x : ClassSplit(name=x, split_label=r[0], class_label=get_label(x)))) \ | |
.map(lambda x : class_split_to_copy_vec(base, x)) | |
copy_vectors.for_each(copy_makedir) | |
def construct_model(): | |
return keras.Sequential([ | |
layers.Conv2D(32, kernel_size=(3, 3), activation='relu',padding='same', | |
input_shape=(IMG_HEIGHT, IMG_WIDTH, 3)), | |
layers.MaxPooling2D(pool_size=(2, 2)), | |
layers.Conv2D(64, kernel_size=(3, 3), activation='relu',padding='same' ), | |
layers.MaxPooling2D(pool_size=(2, 2)), | |
layers.Conv2D(128, kernel_size=(3, 3), activation='relu',padding='same'), | |
layers.MaxPooling2D(pool_size=(2, 2)), | |
layers.Flatten(), | |
# feed-forward classifier | |
layers.Dense(256, activation='relu'), | |
layers.Dropout(0.5), | |
layers.Dense(1, activation='sigmoid') | |
]) | |
def summarise_plot(res): | |
epochs = range(0, len(res.history['val_acc'])) | |
#plot loss | |
plt.subplot(211) | |
plt.plot(res.history['loss'], color='blue',marker='o', label='Training Loss') | |
plt.plot(res.history['val_loss'], color='red',marker='o', label='Validation Loss') | |
plt.xlabel("Epochs") | |
plt.ylabel("Loss") | |
plt.legend() | |
# plot accuracy | |
plt.subplot(212) | |
plt.plot(epochs, res.history['acc'], marker='o', color='red', label="Training Accuracy") | |
plt.plot(epochs, res.history['val_acc'], marker='o', color='blue', label="Validation Accuracy") | |
plt.xlabel("Epochs") | |
plt.ylabel("Accuracy") | |
plt.legend() | |
# save plot to file | |
plt.savefig('pet_class_plot_{}.png'.format(time())) | |
plt.close() | |
def main(): | |
model = construct_model() | |
model.compile(optimizer=optimizers.Adam(), loss='binary_crossentropy', metrics=['accuracy']) | |
model.summary() | |
train_data_gen = \ | |
ImageDataGenerator(rotation_range=40, | |
width_shift_range=0.1, | |
height_shift_range=0.1, | |
rescale=1.0/255, | |
shear_range=0.1, | |
zoom_range=0.1, | |
horizontal_flip=True, | |
fill_mode='nearest') | |
test_data_gen = ImageDataGenerator(rescale=1.0/255) | |
train_iter = train_data_gen.flow_from_directory('dataset/training/', shuffle=True, | |
class_mode='binary', batch_size=BATCH_SIZE, target_size=(IMG_HEIGHT, IMG_WIDTH)) | |
validation_iter = test_data_gen.flow_from_directory('dataset/validation/', shuffle=True, | |
class_mode='binary', batch_size=BATCH_SIZE, target_size=(IMG_HEIGHT, IMG_WIDTH)) | |
test_iter = test_data_gen.flow_from_directory('dataset/test/', shuffle=True, | |
class_mode='binary', batch_size=BATCH_SIZE, target_size=(IMG_HEIGHT, IMG_WIDTH)) | |
res = model.fit_generator(train_iter, steps_per_epoch=len(train_iter), | |
validation_data=validation_iter, validation_steps=len(validation_iter), | |
epochs=EPOCHS, verbose=1, workers=4) | |
loss, acc = model.evaluate_generator(test_iter, steps=len(test_iter), verbose=0) | |
print('Test: [ loss = {}; accuracy = {} ]'.format(loss, acc)) | |
model.save('pet_classifier_ccn.h5') | |
summarise_plot(res) | |
if __name__ == '__main__': | |
main() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment