TDSBirdClassification2.py

import numpy as np
import os
import matplotlib.pyplot as plt
import seaborn as sns

from numpy.random import seed
seed(1337)
from tensorflow import set_random_seed
set_random_seed(42)

from tensorflow.python.keras.applications import vgg16
from tensorflow.python.keras.applications.vgg16 import preprocess_input
from tensorflow.python.keras.preprocessing.image import ImageDataGenerator, load_img
from tensorflow.python.keras.callbacks import ModelCheckpoint
from tensorflow.python.keras import layers, models, Model, optimizers

from sklearn.metrics import classification_report, confusion_matrix, accuracy_score
from plot_conf_matr import plot_confusion_matrix

train_data_dir = "data/train"
val_data_dir = "data/val"
test_data_dir = "data/test"
category_names = sorted(os.listdir('data/train'))
nb_categories = len(category_names)
img_pr_cat = []
for category in category_names:
    folder = 'data/train' + '/' + category
    img_pr_cat.append(len(os.listdir(folder)))
sns.barplot(y=category_names, x=img_pr_cat).set_title("Number of training images per category:")

for subdir, dirs, files in os.walk('data/train'):
    for file in files:
        img_file = subdir + '/' + file
        image = load_img(img_file)
        plt.figure()
        plt.title(subdir)
        plt.imshow(image)
        break

img_height, img_width = 224,224
conv_base = vgg16.VGG16(weights='imagenet', include_top=False, pooling='max', input_shape = (img_width, img_height, 3))

model = models.Sequential()
model.add(conv_base)
model.add(layers.Dense(nb_categories, activation='softmax'))
model.summary()

#Number of images to load at each iteration
batch_size = 32
# only rescaling
train_datagen =  ImageDataGenerator(
    rescale=1./255
)
test_datagen =  ImageDataGenerator(
    rescale=1./255
)
# these are generators for train/test data that will read pictures #found in the defined subfolders of 'data/'
print('Total number of images for "training":')
train_generator = train_datagen.flow_from_directory(
    train_data_dir,
    target_size = (img_height, img_width),
    batch_size = batch_size,
    class_mode = "categorical")
print('Total number of images for "validation":')
val_generator = test_datagen.flow_from_directory(
    val_data_dir,
    target_size = (img_height, img_width),
    batch_size = batch_size,
    class_mode = "categorical",
    shuffle=False)
print('Total number of images for "testing":')
test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size = (img_height, img_width),
    batch_size = batch_size,
    class_mode = "categorical",
    shuffle=False)

learning_rate = 5e-5
epochs = 10
checkpoint = ModelCheckpoint("sign_classifier.h5", monitor = 'val_acc', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)
model.compile(loss="categorical_crossentropy", optimizer=optimizers.Adam(lr=learning_rate, clipnorm = 1.), metrics = ['acc'])

history = model.fit_generator(train_generator,
                              epochs=epochs,
                              shuffle=True,
                              validation_data=val_generator,
                              callbacks=[checkpoint]
                              )

model = models.load_model("sign_classifier.h5")

accuracy = accuracy_score(test_generator.classes, y_pred)
print("Accuracy in test set: %0.1f%% " % (accuracy * 100))

# Image Augmentation
conv_base = vgg16.VGG16(weights='imagenet', include_top=False, pooling='max', input_shape = (img_width, img_height, 3))

#for layer in conv_base.layers[:-13]:
#    layer.trainable = False
model = models.Sequential()
model.add(conv_base)
model.add(layers.Dense(nb_categories, activation='softmax'))

train_datagen = ImageDataGenerator(
        rescale=1./255,
        rotation_range=10,
        zoom_range=0.1,
        width_shift_range=0.1,
        height_shift_range=0.1,
        horizontal_flip=False,
        brightness_range = (0.9,1.1),
        fill_mode='nearest'
        )

# this is a generator that will read pictures found in
# subfolers of 'data/train', and indefinitely generate
# batches of augmented image data

train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size = (img_height, img_width),
batch_size = batch_size,
#save_to_dir='augm_images',
save_prefix='aug',
save_format='jpg',
class_mode = "categorical")

learning_rate = 5e-5
epochs = 20
checkpoint = ModelCheckpoint("sign_classifier_augm.h5", monitor='val_acc', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)
model.compile(loss="categorical_crossentropy", optimizer=optimizers.Adam(lr=learning_rate, clipnorm=1.), metrics = ['acc'])
history = model.fit_generator(train_generator,
                              epochs=epochs,
                              shuffle=True,
                              validation_data=test_generator,
                              callbacks=[checkpoint]
                              )