In [1]:

    

import tensorflow as tf
from PIL import Image
import numpy as np


    

In [2]:

    

# variables

img_size = 28

pos_num = 100
neg_num = 100

train_x = []
train_y = []

for i in range(pos_num):
    with Image.open("/Users/sungju/Dropbox/data/pos_%d.jpg"%(i)) as img:
        img = img.resize((img_size,img_size))
        img = np.array(list(img.getdata()))
        img = img.reshape((img_size,img_size,3))
        train_x.append(img)
        train_y.append([0, 1])

for i in range(neg_num):
    with Image.open("/Users/sungju/Dropbox/data/neg_%d.jpg"%(i)) as img:
        img = img.resize((img_size,img_size))
        img = np.array(list(img.getdata()))
        img = img.reshape((img_size,img_size,3))
        train_x.append(img)
        train_y.append([1, 0])

train_x = np.array(train_x)
train_y = np.array(train_y)

perm = np.random.permutation(len(train_y))

train_x = train_x[perm]
train_y = train_y[perm]


    

In [3]:

    

train_x.shape


    

    
Out[3]:





(200, 28, 28, 3)

In [4]:

    

train_y.shape


    

    
Out[4]:





(200, 2)

In [5]:

    

learning_rate = 0.01
epochs = 10000
batch_size = 8
display_step = 10

n_channel = 3
n_classes = 2

dropout = 0.75

X = tf.placeholder(tf.float32, [None, img_size, img_size, n_channel])
Y = tf.placeholder(tf.float32, [None, n_classes])
keep_prob = tf.placeholder(tf.float32)


    

In [6]:

    

def conv2d(X, W, b, strides=1):
    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):
    # ksize is window!, strides is just strides!
    return tf.nn.max_pool(X,ksize = [1,k,k,1],strides=[1,k,k,1], padding='SAME')

def conv_net(X, weight, biases, dropout):
    X = tf.reshape(X, shape=[-1, img_size, img_size, n_channel])
    
    conv1 = conv2d(X, weights['wc1'], biases['bc1'])
    conv1 = maxpool2d(conv1,k=2)
    
    conv2 = conv2d(conv1, weights['wc2'], biases['bc2'])
    conv2 = maxpool2d(conv2, k=2)
    
    fc1 = tf.reshape(conv2, [-1, weights['wd1'].get_shape().as_list()[0]])
    print(weights['wd1'].get_shape().as_list()[0])
    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])
    fc1 = tf.nn.relu(fc1)
    
    fc1 = tf.nn.dropout(fc1, dropout)
    
    out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])
    return out


    

In [7]:

    

weights = {
    # 5x5 conv, 1 input, 32 outputs
    # I think '1 input' means the channel's dimension is one
    'wc1': tf.Variable(tf.random_normal([5,5,3,32])),
    'wc2': tf.Variable(tf.random_normal([5,5,32,64])),
    'wd1': tf.Variable(tf.random_normal([7*7*64, 1024])),
    'out': tf.Variable(tf.random_normal([1024, n_classes]))
}

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


    

In [8]:

    

pred = conv_net(X,weights, biases, keep_prob)

cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(pred,Y))
optimizer = tf.train.AdamOptimizer(learning_rate).minimize(cost)

correct_pred = tf.equal(tf.argmax(pred,1), tf.argmax(Y,1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

init = tf.initialize_all_variables()


    

    




3136

In [9]:

    

with tf.Session() as sess:
    sess.run(init)
    step = 1
    
    while step* batch_size < epochs:
        
        offset = (step * batch_size) % (train_y.shape[0] - batch_size)
        batch_x = train_x[offset:(offset + batch_size), :, :, :]
        batch_y = train_y[offset:(offset + batch_size), :]
        
        sess.run(optimizer, feed_dict={X:batch_x, Y:batch_y, keep_prob:dropout})
        
        if step% display_step == 0:
            loss, acc = sess.run([cost,accuracy], feed_dict={X:batch_x, Y:batch_y,
                                                               keep_prob: 1.})
            
            print( "Iter " + str(step*batch_size) + ", Minibatch Loss= " + \
                  "{:.6f}".format(loss) + ", Training Accuracy= " + \
                  "{:.5f}".format(acc))
            
        step +=1
        
    print("Optimization Finished!")
    print("Testing Accuracy"+ str(sess.run(accuracy, feed_dict={X: train_x,
                                                                    Y: train_y,
                                                                      keep_prob:1.})))


    

    




Iter 80, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 160, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 240, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 320, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 400, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 480, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 560, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 640, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 720, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 800, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 880, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 960, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1040, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1120, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1200, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1280, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1360, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1440, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1520, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1600, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1680, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1760, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1840, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 1920, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2000, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2080, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2160, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2240, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2320, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2400, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2480, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2560, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2640, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2720, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2800, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2880, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 2960, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3040, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3120, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3200, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3280, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3360, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3440, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3520, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3600, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3680, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3760, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3840, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 3920, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4000, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4080, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4160, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4240, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4320, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4400, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4480, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4560, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4640, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4720, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4800, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4880, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 4960, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5040, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5120, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5200, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5280, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5360, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5440, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5520, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5600, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5680, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5760, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5840, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 5920, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6000, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6080, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6160, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6240, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6320, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6400, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6480, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6560, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6640, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6720, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6800, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6880, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 6960, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7040, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7120, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7200, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7280, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7360, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7440, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7520, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7600, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7680, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7760, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7840, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 7920, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8000, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8080, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8160, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8240, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8320, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8400, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8480, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8560, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8640, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8720, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8800, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8880, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 8960, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9040, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9120, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9200, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9280, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9360, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9440, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9520, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9600, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9680, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9760, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9840, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Iter 9920, Minibatch Loss= 0.000000, Training Accuracy= 1.00000
Optimization Finished!
Testing Accuracy1.0

In [ ]: