VGG_16_car_logo_detector.py

from keras.layers import Input, Lambda, Dense, Flatten
from keras.models import Model
from keras.applications.vgg16 import VGG16
from keras.applications.vgg16 import preprocess_input
from keras.preprocessing import image
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
import numpy as np
from glob import glob
import matplotlib.pyplot as plt

# The default image dimension in VGG-16
IMAGE_SIZE = [224, 224]

# Add addresses where the dataset is stored
train_path = 'D:/DATA_BASE/cars_logo/Train'
valid_path = 'D:/DATA_BASE/cars_logo/Test'

# Pre-processing layer for VGG-16
# If we don't set include_top as False then we'll have a model which classifies a thousand categories when we only have 10 i.e Disaster
vgg = VGG16(input_shape=IMAGE_SIZE + [3], weights='imagenet', include_top=False)

# Turn OFF training for layers
for layer in vgg.layers:
  layer.trainable = False

folders = glob('D:/DATA_BASE/cars_logo/Train/*')

print(folders)

# Flattening the feature map
x = Flatten()(vgg.output)
# x = Dense(1000, activation='relu')(x)
# Adding the last layer with number of neurons = number of categories
prediction = Dense(len(folders), activation='softmax')(x)

# create a model object
model = Model(inputs=vgg.input, outputs=prediction)

# view the structure of the model
model.summary()

# tell the model what cost and optimization method to use
model.compile(
  loss='categorical_crossentropy',
  optimizer='adam',
  metrics=['accuracy']
)


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('D:/DATA_BASE/cars_logo/Train',
                                                 target_size = (224, 224),
                                                 batch_size = 32,
                                                 class_mode = 'categorical')

test_set = test_datagen.flow_from_directory('D:/DATA_BASE/cars_logo/Test',
                                            target_size = (224, 224),
                                            batch_size = 32,
                                            class_mode = 'categorical')

# fit the model
r = model.fit_generator(
  training_set,
  validation_data=test_set,
  epochs=5,
  steps_per_epoch=len(training_set),
  validation_steps=len(test_set)
)
# loss
plt.plot(r.history['loss'], label='train loss')
plt.plot(r.history['val_loss'], label='val loss')
plt.legend()
plt.show()
plt.savefig('LossVal_loss')

# accuracies
plt.plot(r.history['accuracy'], label='train acc')
plt.plot(r.history['val_accuracy'], label='val acc')
plt.legend()
plt.show()
plt.savefig('AccVal_acc')

import tensorflow as tf

from keras.models import load_model

model.save('VGG16_Car_Logo_Model.h5')