train.py

from __future__ import print_function
import tensorflow as tf
import numpy as np
from PIL import Image
import json


# Parameters
learning_rate = 0.001
training_iters = 3000
batch_size = 20
display_step = 1

# Network Parameters
n_input = 128*128 # Cropped Image data input (img shape: 128*128 )
n_classes = 10 # Total classes (0-9 digits)
dropout = 0.75 # Dropout, probability to keep units

# tf Graph input
x = tf.placeholder(tf.float32, [None, 128, 128, 3])
y = tf.placeholder(tf.float32, [None, n_classes])
keep_prob = tf.placeholder(tf.float32) #dropout (keep probability)

#Cropped Images Directory
CROPPED_IMG_DIC = './cropped_img/'

# Store layers weight & bias
weights = {
    # 5x5 conv, 3 input, 16 outputs
    'wc1': tf.Variable(tf.random_normal([5, 5, 3, 16])),
    # 5x5 conv, 16 inputs, 32 outputs
    'wc2': tf.Variable(tf.random_normal([5, 5, 16, 32])),
    # 5x5 conv, 32 inputs, 64 outputs
    'wc3': tf.Variable(tf.random_normal([5, 5, 32 ,64])),
    # fully connected, 32*32*96 inputs, 1024 outputs
    'wd1': tf.Variable(tf.random_normal([16*16*64, 1024])),
    # 1024 inputs, 10 outputs (class prediction)
    'out': tf.Variable(tf.random_normal([1024, n_classes]))
}

biases = {
    'bc1': tf.Variable(tf.random_normal([16])),
    'bc2': tf.Variable(tf.random_normal([32])),
    'bc3': tf.Variable(tf.random_normal([64])),
    'bd1': tf.Variable(tf.random_normal([1024])),
    'out': tf.Variable(tf.random_normal([n_classes]))
}


def conv2d(x, W, b, strides=1):
    # Conv2D wrapper, with bias and relu activation
    x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding='SAME')
    x = tf.nn.bias_add(x, b)
    return tf.nn.relu(x)


def maxpool2d(x, k=2):
    # MaxPool2D wrapper
    return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1],
                          padding='SAME')

def conv_net(x, weights, biases, dropout):
    #Reshape input picture
    x = tf.reshape(x, shape=[-1, 128, 128, 3])

    #Conv Layer and ReLU #1
    conv1 = conv2d(x, weights['wc1'], biases['bc1'])
    print(conv1.shape)

    #Max Pool #1
    mp1 = maxpool2d(conv1, k = 2)

    #Conv Layer and ReLU #2
    conv2 = conv2d(mp1, weights['wc2'], biases['bc2'])
    print(conv2.shape)

    #Max Pool #2
    mp2 = maxpool2d(conv2, k = 2)

    #Conv Layer and ReLU #3
    conv3 = conv2d(mp2, weights['wc3'], biases['bc3'])
    print(conv3.shape)

    #Max Pool #3
    mp3 = maxpool2d(conv3, k = 2)

    #Fully Connected Neural Network #1
    fc1 = tf.reshape(mp3, [-1, weights['wd1'].get_shape().as_list()[0]])
    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])
    fc1 = tf.nn.relu(fc1)

    #Dropout Layer
    fc1 = tf.nn.dropout(fc1, dropout)

    #Fully Connected Neural Network #2 Output Layer
    out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])
    return out





# Construct model
pred = conv_net(x, weights, biases, keep_prob)

# Define loss and optimizer
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)

# Evaluate model
correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

# Initializing the variables
init = tf.global_variables_initializer()
saver=tf.train.Saver()

# Launch the graph
with tf.Session() as sess:
    sess.run(init)
    step = 1
    # Keep training until reach max iterations

    #read tags from json file
    tags = {}
    with open('tags.json', 'r') as json_file:
        json_str = json_file.read()
        tags = json.loads(json_str)

    tagged_imgs = []
    for fname in tags:
        t = (fname, tags[fname])
        tagged_imgs.append(t)

    for i in range(0,10):
        num_batch = len(tagged_imgs) // batch_size
        print(num_batch)
        for batch_id in range(0, num_batch):
            start_index = batch_id * batch_size
            end_index = (batch_id + 1) * batch_size
            batch = tagged_imgs[start_index: end_index]

            #build batch
            batch_xs = []
            batch_ys = []
            for fname, tags in batch:
                img = Image.open(CROPPED_IMG_DIC + fname)
                img_ndarray = np.asarray(img, dtype='float32')
                img_ndarray = np.reshape(img_ndarray, [128, 128, 3])
                batch_xs.append(img_ndarray)

                #Set tags
                batch_y = np.zeros(n_classes)
                batch_y[tags - 1] = 1
                batch_y = np.reshape(batch_y, [n_classes, ])
                batch_ys.append(batch_y)
            batch_xs = np.asarray(batch_xs)
            batch_ys = np.asarray(batch_ys)
            sess.run(optimizer, feed_dict={x: batch_xs, y: batch_ys,
                                           keep_prob: dropout})
            if step % display_step == 0:
                # Calculate batch loss and accuracy
                loss, acc = sess.run([cost, accuracy], feed_dict={x: batch_xs,
                                                                  y: batch_ys,
                                                                  keep_prob: 1.})
                print("Iter " + str(step*batch_size) + ", Minibatch Loss= " + \
                      "{:.6f}".format(loss) + ", Training Accuracy= " + \
                      "{:.5f}".format(acc))
            step += 1
    #Split to batches
    num_batch = len(tagged_imgs) // batch_size
    print(num_batch)
    for batch_id in range(0, num_batch):
        start_index = batch_id * batch_size
        end_index = (batch_id + 1) * batch_size
        batch = tagged_imgs[start_index: end_index]

        #build batch
        batch_xs = []
        batch_ys = []
        for fname, tags in batch:
            img = Image.open(CROPPED_IMG_DIC + fname)
            img_ndarray = np.asarray(img, dtype='float32')
            img_ndarray = np.reshape(img_ndarray, [128, 128, 3])
            batch_xs.append(img_ndarray)

            #Set tags
            batch_y = np.zeros(n_classes)
            batch_y[tags - 1] = 1
            batch_y = np.reshape(batch_y, [n_classes, ])
            batch_ys.append(batch_y)
        batch_xs = np.asarray(batch_xs)
        batch_ys = np.asarray(batch_ys)
        sess.run(optimizer, feed_dict={x: batch_xs, y: batch_ys,
                                       keep_prob: dropout})
        if step % display_step == 0:
            # Calculate batch loss and accuracy
            loss, acc = sess.run([cost, accuracy], feed_dict={x: batch_xs,
                                                              y: batch_ys,
                                                              keep_prob: 1.})
            print("Iter " + str(step*batch_size) + ", Minibatch Loss= " + \
                  "{:.6f}".format(loss) + ", Training Accuracy= " + \
                  "{:.5f}".format(acc))
        step += 1
    print("Optimization Finished!")
    saver.save(sess,"./model.ckpt")