Train.py

# coding: utf-8

# In[1]:

import tensorflow as tf
from keras.backend.tensorflow_backend import set_session
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
set_session(tf.Session(config=config))



batch_size = 10
epochs_num = 20
height = 192
weight = 168


import os
import csv

path = os.getcwd()

data_path = '/Tooth_Brushing_Dataset/image_data/train.csv'

Data_path = []

with open(path+data_path, newline='') as csvfile:
    spamreader = csv.reader(csvfile, delimiter=' ', quotechar='|')
    for row in spamreader:
        x,y = str(row[0]).split(',')
        Data_path.append('{} {}\n'.format(path+'/Tooth_Brushing_Dataset/image_data/train/'+str(x)+'.jpg', y))
        
Train_file = open('Train_list.txt','w')
Test_file = open('Test_list.txt', 'w')

Train_num = 6000
Test_num = 1316

print(len(Data_path))

for i in range(len(Data_path)):
    if i<Train_num:
        Train_file.write(Data_path[i])
    else:
        Test_file.write(Data_path[i])

Train_file.close()
Test_file.close()


# In[2]:

import numpy as np
from PIL import Image

def path2img(path):
    im = Image.open(path)
    img = np.array(im, 'uint8')
    return img


# In[3]:

from keras.utils import to_categorical
from keras.preprocessing.image import ImageDataGenerator


def gen_sample_from_file(path, batch_size):
    while True:
        f = open(path)
        Batch_x = []
        Batch_y = []
        cnt = 0
        for line in f:
            cnt = cnt + 1
            if cnt % batch_size == 0:
                Batch_x = np.array(Batch_x, dtype=np.float32)
                Batch_y = np.array(Batch_y, dtype=np.int32)
                std = ImageDataGenerator(featurewise_center=True, featurewise_std_normalization=True)
                std.fit(Batch_x)
                Batch_x = std.standardize(Batch_x)
                yield (Batch_x, Batch_y)
                Batch_x = []
                Batch_y = []
            img_path, y = line.split()
            im = Image.open(img_path)
            im = im.resize([height, weight])
            img = np.array(im, 'uint')
            img = np.reshape(img, [height, weight, 3])
            y = int(y)-1
            y = to_categorical(y, num_classes=16)
            y = np.reshape(y, [16])
            Batch_x.append(img) 
            Batch_y.append(y)
        f.close()



# In[ ]:

# In[ ]:

from keras.models import Sequential
from keras.layers import *
from keras.optimizers import SGD

model = Sequential()
model.add(Conv2D(96, kernel_size=(11, 11), activation='relu', input_shape=(height, weight, 3)))
model.add(MaxPooling2D(pool_size=(3, 3)))
model.add(Conv2D(256, kernel_size=(5, 5), activation='relu'))
model.add(MaxPooling2D(pool_size=(3, 3)))
model.add(Conv2D(384, kernel_size=(5, 5)))
model.add(Conv2D(384, kernel_size=(5, 5)))
model.add(Conv2D(256, kernel_size=(5, 5)))
model.add(MaxPooling2D(pool_size=(3, 3)))
model.add(Flatten())
model.add(Dense(1024))
model.add(Dropout(0.3))
model.add(Dense(1024))
model.add(Dense(16, activation='softmax'))

sgd = SGD(lr=0.001, decay=0.01, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
model.summary()
hist = model.fit_generator(generator = gen_sample_from_file('Train_list.txt', batch_size),validation_data= gen_sample_from_file('Test_list.txt', batch_size), validation_steps = 10 , steps_per_epoch = 600, epochs = epochs_num)
model.save_weights('my_model_weights.h5')

from keras.backend import clear_session
clear_session()
# In[ ]:

f = open('lossfile.txt','w')
f.write('Loss')
for i in hist.history['loss']:
    f.write(str(i)+'\n')
f.write('val_loss')
for i in hist.history['val_loss']:
    f.write(str(i)+'\n')
f.write('acc')
for i in hist.history['acc']:
    f.write(str(i)+'\n')
f.write('val_acc')
for i in hist.history['val_acc']:
    f.write(str(i)+'\n')
f.close()