├── .ipynb_checkpoints
└── Soft_Attention_MNIST-checkpoint.ipynb
├── Hard_Attention_MNIST.ipynb
├── Images
├── Attention_for_Image_Caption.png
├── Hard_result.png
├── MNIST_sample.png
├── Soft_result.png
├── hard_attention.png
├── soft_attention.png
└── 그림자료.pptx
├── README.md
├── Soft_Attention_MNIST.ipynb
└── saved_networks
└── checkpoint
/Images/Attention_for_Image_Caption.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/Kyushik/Attention/9244708224556c58c8a5741042771a9a46cbda59/Images/Attention_for_Image_Caption.png
--------------------------------------------------------------------------------
/Images/Hard_result.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/Kyushik/Attention/9244708224556c58c8a5741042771a9a46cbda59/Images/Hard_result.png
--------------------------------------------------------------------------------
/Images/MNIST_sample.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/Kyushik/Attention/9244708224556c58c8a5741042771a9a46cbda59/Images/MNIST_sample.png
--------------------------------------------------------------------------------
/Images/Soft_result.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/Kyushik/Attention/9244708224556c58c8a5741042771a9a46cbda59/Images/Soft_result.png
--------------------------------------------------------------------------------
/Images/hard_attention.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/Kyushik/Attention/9244708224556c58c8a5741042771a9a46cbda59/Images/hard_attention.png
--------------------------------------------------------------------------------
/Images/soft_attention.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/Kyushik/Attention/9244708224556c58c8a5741042771a9a46cbda59/Images/soft_attention.png
--------------------------------------------------------------------------------
/Images/그림자료.pptx:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/Kyushik/Attention/9244708224556c58c8a5741042771a9a46cbda59/Images/그림자료.pptx
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # Attention
2 |
3 | ## Introduction
4 | This repository is for algorithms of `Attention`.
5 |
6 | The paper I implemented is as follows.
7 |
8 | - [Show, Attend and Tell: Neural Image Caption Generation with Visual Attention](https://arxiv.org/abs/1502.03044)
9 |
10 |
11 |
12 | ### Dataset
13 |
14 | This Algorithm will be tested by `Modified MNIST dataset` Which is made by [Jongwon Park](https://github.com/jwpark116).
15 |
16 | This modified MNIST dataset is good for verifying attention algorithm.
17 |
18 | The example of modified MNIST is as follows.
19 |
20 | 
21 |
22 | You can download this modified MNIST data from this link
23 |
24 | [Training dataset](https://www.dropbox.com/s/e7jxyulxx2anqyq/MNIST_data_train_re.mat?dl=0) / [Testing dataset](https://www.dropbox.com/s/fcw4754bavva9my/MNIST_data_test_re.mat?dl=0)
25 |
26 |
27 |
28 | ### Environment
29 | **Software**
30 | * Windows7 (64bit)
31 | * Python 3.5.2
32 | * Anaconda 4.2.0
33 | * Tensorflow-gpu 1.4.0
34 |
35 | **Hardware**
36 | * CPU: Intel(R) Core(TM) i7-4790K CPU @ 4.00GHZ
37 | * GPU: GeForce GTX 1080
38 | * Memory: 8GB
39 |
40 |
41 |
42 | ## Algorithms
43 |
44 | ### Soft Attention
45 |
46 | This algorithm is from the paper [Show, Attend and Tell: Neural Image Caption Generation with Visual Attention](https://arxiv.org/abs/1502.03044). I studied attention from [Heuritech blog](https://blog.heuritech.com/2016/01/20/attention-mechanism/).
47 |
48 | The attention model for image captioning from paper is as follows. The image is from the [Heuritech blog](https://blog.heuritech.com/2016/01/20/attention-mechanism/).
49 |
50 | 
51 |
52 | For implementing this algorithm, `Attention model` and `LSTM` are needed. The code of LSTM is as follows.
53 |
54 | ```python
55 | # LSTM function
56 | def LSTM_cell(C_prev, h_prev, x_lstm, Wf, Wi, Wc, Wo, bf, bi, bc, bo):
57 | # C_prev: Cell state from lstm of previous time step (shape: [batch_size, lstm_size])
58 | # h_prev: output from lstm of previous time step (shape: [batch_size, lstm_size])
59 | # x_lstm: input of lstm (shape: [batch_size, data_flatten_size])
60 |
61 | input_concat = tf.concat([x_lstm, h_prev], 1)
62 | f = tf.sigmoid(tf.matmul(input_concat, Wf) + bf)
63 | i = tf.sigmoid(tf.matmul(input_concat, Wi) + bi)
64 | c = tf.tanh(tf.matmul(input_concat, Wc) + bc)
65 | o = tf.sigmoid(tf.matmul(input_concat, Wo) + bo)
66 |
67 | C_t = tf.multiply(f, C_prev) + tf.multiply(i, c)
68 | h_t = tf.multiply(o, tf.tanh(C_t))
69 |
70 | return C_t, h_t # Cell state, Output
71 | ```
72 |
73 | [Colah's blog post](http://colah.github.io/posts/2015-08-Understanding-LSTMs/) is very good for understanding LSTM and I studied this post to implement LSTM.
74 |
75 | Structure image of soft attention model is as follows. Image is from [Heuritech blog](https://blog.heuritech.com/2016/01/20/attention-mechanism/).
76 |
77 | 
78 |
79 | Also, the code of soft attention is as follows.
80 |
81 | ```python
82 | # Soft Attention function
83 | def soft_attention(h_prev, a, Wa, Wh):
84 | # h_prev: output from lstm of previous time step (shape: [batch_size, lstm_size])
85 | # a: Result of CNN [batch_size, conv_size * conv_size, channel_size]
86 |
87 | m_list = [tf.tanh(tf.matmul(a[i], Wa) + tf.matmul(h_prev, Wh)) for i in range(len(a))]
88 | m_concat = tf.concat([m_list[i] for i in range(len(a))], axis = 1)
89 | alpha = tf.nn.softmax(m_concat)
90 | z_list = [tf.multiply(a[i], tf.slice(alpha, (0, i), (-1, 1))) for i in range(len(a))]
91 | z_stack = tf.stack(z_list, axis = 2)
92 | z = tf.reduce_sum(z_stack, axis = 2)
93 |
94 | return alpha, z
95 | ```
96 |
97 | After 10 epoch, The training accuracy of LSTM was 94% and validation accuracy was 97%.
98 |
99 | Sample images of soft attention result are as follows.
100 |
101 | 
102 |
103 |
104 |
105 | ### Hard Attention
106 |
107 | This algorithm is from the paper [Show, Attend and Tell: Neural Image Caption Generation with Visual Attention](https://arxiv.org/abs/1502.03044). Hard Attention architecture image from [Heuritech blog](https://blog.heuritech.com/2016/01/20/attention-mechanism/) is as follows.
108 |
109 | 
110 |
111 | The random choice algorithm is `Monte-Carlo Sampling`. Therefore, I made a code for hard attention as follows.
112 |
113 | ```python
114 | # Hard Attention function
115 | def hard_attention(h_prev, a, Wa, Wh):
116 | # h_prev: output from lstm of previous time step (shape: [batch_size, lstm_size])
117 | # a: Result of CNN [batch_size, conv_size * conv_size, channel_size]
118 |
119 | m_list = [tf.tanh(tf.matmul(a[i], Wa) + tf.matmul(h_prev, Wh)) for i in range(len(a))]
120 | m_concat = tf.concat([m_list[i] for i in range(len(a))], axis = 1)
121 | alpha = tf.nn.softmax(m_concat)
122 |
123 | #For Monte-Carlo Sampling
124 | alpha_cumsum = tf.cumsum(alpha, axis = 1)
125 | len_batch = tf.shape(alpha_cumsum)[0]
126 | rand_prob = tf.random_uniform(shape = [len_batch, 1], minval = 0., maxval = 1.)
127 | alpha_relu = tf.nn.relu(rand_prob - alpha_cumsum)
128 | alpha_index = tf.count_nonzero(alpha_relu, 1)
129 | alpha_hard = tf.one_hot(alpha_index, len(a))
130 |
131 | z_list = [tf.multiply(a[i], tf.slice(alpha_hard, (0, i), (-1, 1))) for i in range(len(a))]
132 | z_stack = tf.stack(z_list, axis = 2)
133 | z = tf.reduce_sum(z_stack, axis = 2)
134 |
135 | return alpha, z
136 | ```
137 |
138 |
139 |
140 | After 10 epoch, The training accuracy of LSTM was only 30% and validation accuracy was 33%.
141 |
142 | Sample images of hard attention result are as follows.
143 |
144 | 
145 |
146 |
147 |
148 |
--------------------------------------------------------------------------------
/Soft_Attention_MNIST.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {
6 | "collapsed": true
7 | },
8 | "source": [
9 | "# Soft Attention MNIST\n",
10 | "\n",
11 | "This is jupyter notebook for `Soft Attention` from paper [Show, Attend and Tell](https://arxiv.org/abs/1502.03044). \n",
12 | "
This Algorithm will be tested by `Modified MNIST dataset` Which is made by [Jongwon Park](https://github.com/jwpark116).
This modified MNIST dataset is good for verifying attention algorithm.\n",
13 | "
You can download modified MNIST data from this link\n",
14 | "
[Training dataset](https://www.dropbox.com/s/e7jxyulxx2anqyq/MNIST_data_train_re.mat?dl=0) / [Testing dataset](https://www.dropbox.com/s/fcw4754bavva9my/MNIST_data_test_re.mat?dl=0)"
15 | ]
16 | },
17 | {
18 | "cell_type": "code",
19 | "execution_count": 1,
20 | "metadata": {
21 | "scrolled": false
22 | },
23 | "outputs": [
24 | {
25 | "name": "stderr",
26 | "output_type": "stream",
27 | "text": [
28 | "/opt/anaconda/lib/python3.6/site-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
29 | " from ._conv import register_converters as _register_converters\n"
30 | ]
31 | }
32 | ],
33 | "source": [
34 | "# Import modules\n",
35 | "import tensorflow as tf\n",
36 | "import matplotlib.pyplot as plt\n",
37 | "import numpy as np\n",
38 | "import scipy.io\n",
39 | "import random\n",
40 | "import skimage.transform"
41 | ]
42 | },
43 | {
44 | "cell_type": "code",
45 | "execution_count": 2,
46 | "metadata": {},
47 | "outputs": [
48 | {
49 | "name": "stdout",
50 | "output_type": "stream",
51 | "text": [
52 | "Train data shape: (55000, 112, 112)\n",
53 | "Train label shape: (55000, 10)\n",
54 | "Test data shape: (9900, 112, 112)\n",
55 | "Test label shape: (9900, 10)\n",
56 | "Validation data shape: (100, 112, 112)\n",
57 | "Validation label shape: (100, 10)\n"
58 | ]
59 | }
60 | ],
61 | "source": [
62 | "# Import MNIST data\n",
63 | "\n",
64 | "mat_data_train = scipy.io.loadmat('MNIST_data_train_re.mat')\n",
65 | "mat_data_test = scipy.io.loadmat('MNIST_data_test_re.mat')\n",
66 | "\n",
67 | "train_x = mat_data_train['X_train']\n",
68 | "train_y = mat_data_train['Y_train']\n",
69 | "\n",
70 | "test_x = mat_data_test['X_test'][:9900, :]\n",
71 | "test_y = mat_data_test['Y_test'][:9900, :]\n",
72 | "\n",
73 | "validation_x = mat_data_test['X_test'][9900:, :]\n",
74 | "validation_y = mat_data_test['Y_test'][9900:, :]\n",
75 | "\n",
76 | "del mat_data_train\n",
77 | "del mat_data_test\n",
78 | "\n",
79 | "print(\"Train data shape: \" + str(train_x.shape))\n",
80 | "print(\"Train label shape: \" + str(train_y.shape))\n",
81 | "print(\"Test data shape: \" + str(test_x.shape))\n",
82 | "print(\"Test label shape: \" + str(test_y.shape))\n",
83 | "print(\"Validation data shape: \" + str(validation_x.shape))\n",
84 | "print(\"Validation label shape: \" + str(validation_y.shape))"
85 | ]
86 | },
87 | {
88 | "cell_type": "markdown",
89 | "metadata": {},
90 | "source": [
91 | "## Parameters"
92 | ]
93 | },
94 | {
95 | "cell_type": "code",
96 | "execution_count": 3,
97 | "metadata": {
98 | "collapsed": true
99 | },
100 | "outputs": [],
101 | "source": [
102 | "# Parameters \n",
103 | "img_size = train_x.shape[1]\n",
104 | "img_flat_size = img_size * img_size\n",
105 | "\n",
106 | "# If you want to train the model -> True, otherwise -> False\n",
107 | "Is_train = True\n",
108 | "\n",
109 | "# If you want to load saved model -> True, otherwise -> False \n",
110 | "Load_model = False\n",
111 | "\n",
112 | "# Name of the save file\n",
113 | "save_name = 'soft1'\n",
114 | "\n",
115 | "# Numbers of sampling to test the code \n",
116 | "num_test_sample = 10\n",
117 | "\n",
118 | "# labels: 0 - 9\n",
119 | "num_label = 10\n",
120 | "\n",
121 | "# Parameters for training\n",
122 | "num_epoch = 10\n",
123 | "\n",
124 | "learning_rate = 5e-4\n",
125 | "epsilon = 1e-8\n",
126 | "\n",
127 | "batch_size = 256\n",
128 | "\n",
129 | "# Parameter for LSTM\n",
130 | "lstm_size = 256\n",
131 | "step_size = 4\n",
132 | "flatten_size = img_size\n",
133 | "\n",
134 | "gpu_fraction = 0.3"
135 | ]
136 | },
137 | {
138 | "cell_type": "markdown",
139 | "metadata": {},
140 | "source": [
141 | "## Plotting Sample Image (Modified MNIST for Attention)"
142 | ]
143 | },
144 | {
145 | "cell_type": "code",
146 | "execution_count": 4,
147 | "metadata": {},
148 | "outputs": [
149 | {
150 | "data": {
151 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQUAAAD8CAYAAAB+fLH0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
152 | "text/plain": [
153 | ""
154 | ]
155 | },
156 | "metadata": {},
157 | "output_type": "display_data"
158 | },
159 | {
160 | "name": "stdout",
161 | "output_type": "stream",
162 | "text": [
163 | "Label: [0. 0. 0. 0. 0. 0. 0. 1. 0. 0.]\n",
164 | "Shape: 112x112\n"
165 | ]
166 | }
167 | ],
168 | "source": [
169 | "# Plotting example image\n",
170 | "img = train_x[0, :, :]\n",
171 | "\n",
172 | "plt.imshow(img, cmap = 'gray')\n",
173 | "plt.show()\n",
174 | "print('Label: ' + str(train_y[0,:]))\n",
175 | "print('Shape: ' + str(img_size) + 'x' + str(img_size))"
176 | ]
177 | },
178 | {
179 | "cell_type": "markdown",
180 | "metadata": {},
181 | "source": [
182 | "## Functions for Convolutional Network"
183 | ]
184 | },
185 | {
186 | "cell_type": "code",
187 | "execution_count": 5,
188 | "metadata": {
189 | "collapsed": true
190 | },
191 | "outputs": [],
192 | "source": [
193 | "# Initialize weights and bias \n",
194 | "def conv2d(x,w, stride):\n",
195 | "\treturn tf.nn.conv2d(x,w,strides=[1, stride, stride, 1], padding='SAME')\n",
196 | "\n",
197 | "def max_pool_2x2(x):\n",
198 | " return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],\n",
199 | " strides=[1, 2, 2, 1], padding='SAME')\n",
200 | "\n",
201 | "# Get Variables\n",
202 | "def weight_variable(name, shape):\n",
203 | " return tf.get_variable(name,shape=shape, initializer=tf.contrib.layers.xavier_initializer())\n",
204 | "\n",
205 | "def bias_variable(name, shape):\n",
206 | " return tf.get_variable(name,shape=shape, initializer=tf.contrib.layers.xavier_initializer())"
207 | ]
208 | },
209 | {
210 | "cell_type": "markdown",
211 | "metadata": {},
212 | "source": [
213 | "## LSTM and Attention function"
214 | ]
215 | },
216 | {
217 | "cell_type": "code",
218 | "execution_count": 6,
219 | "metadata": {
220 | "collapsed": true
221 | },
222 | "outputs": [],
223 | "source": [
224 | "# Reset the graph\n",
225 | "tf.reset_default_graph()\n",
226 | "\n",
227 | "# LSTM function\n",
228 | "def LSTM_cell(C_prev, h_prev, x_lstm, Wf, Wi, Wc, Wo, bf, bi, bc, bo):\n",
229 | " # C_prev: Cell state from lstm of previous time step (shape: [batch_size, lstm_size])\n",
230 | " # h_prev: output from lstm of previous time step (shape: [batch_size, lstm_size])\n",
231 | " # x_lstm: input of lstm (shape: [batch_size, data_flatten_size])\n",
232 | "\n",
233 | " input_concat = tf.concat([x_lstm, h_prev], 1)\n",
234 | " f = tf.sigmoid(tf.matmul(input_concat, Wf) + bf)\n",
235 | " i = tf.sigmoid(tf.matmul(input_concat, Wi) + bi)\n",
236 | " c = tf.tanh(tf.matmul(input_concat, Wc) + bc)\n",
237 | " o = tf.sigmoid(tf.matmul(input_concat, Wo) + bo)\n",
238 | " \n",
239 | " C_t = tf.multiply(f, C_prev) + tf.multiply(i, c) \n",
240 | " h_t = tf.multiply(o, tf.tanh(C_t))\n",
241 | " \n",
242 | " return C_t, h_t # Cell state, Output\n",
243 | "\n",
244 | "# Soft Attention function\n",
245 | "def soft_attention(h_prev, a, Wa, Wh):\n",
246 | " # h_prev: output from lstm of previous time step (shape: [batch_size, lstm_size])\n",
247 | " # a: Result of CNN [batch_size, conv_size * conv_size, channel_size] \n",
248 | "\n",
249 | " m_list = [tf.tanh(tf.matmul(a[i], Wa) + tf.matmul(h_prev, Wh)) for i in range(len(a))] \n",
250 | " m_concat = tf.concat([m_list[i] for i in range(len(a))], axis = 1) \n",
251 | " alpha = tf.nn.softmax(m_concat) \n",
252 | " z_list = [tf.multiply(a[i], tf.slice(alpha, (0, i), (-1, 1))) for i in range(len(a))]\n",
253 | " z_stack = tf.stack(z_list, axis = 2)\n",
254 | " z = tf.reduce_sum(z_stack, axis = 2)\n",
255 | "\n",
256 | " return alpha, z\n",
257 | " "
258 | ]
259 | },
260 | {
261 | "cell_type": "markdown",
262 | "metadata": {},
263 | "source": [
264 | "## Network"
265 | ]
266 | },
267 | {
268 | "cell_type": "code",
269 | "execution_count": 7,
270 | "metadata": {},
271 | "outputs": [
272 | {
273 | "name": "stdout",
274 | "output_type": "stream",
275 | "text": [
276 | "WARNING:tensorflow:From /opt/anaconda/lib/python3.6/site-packages/tensorflow/contrib/learn/python/learn/datasets/base.py:198: retry (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
277 | "Instructions for updating:\n",
278 | "Use the retry module or similar alternatives.\n"
279 | ]
280 | }
281 | ],
282 | "source": [
283 | "# Network\n",
284 | "\n",
285 | "# Input \n",
286 | "x_image = tf.placeholder(tf.float32, shape = [None, img_size, img_size, 1])\n",
287 | "y_target = tf.placeholder(tf.float32, shape=[None, num_label])\n",
288 | "\n",
289 | "# Convolution variables\n",
290 | "w_conv1 = weight_variable('W_conv1', [3, 3, 1, 64])\n",
291 | "b_conv1 = bias_variable('b_conv1', [64])\n",
292 | "w_conv2 = weight_variable('W_conv2', [3, 3, 64, 256])\n",
293 | "b_conv2 = bias_variable('b_conv2', [256])\n",
294 | "w_conv3 = weight_variable('W_conv3', [3, 3, 256, 512])\n",
295 | "b_conv3 = bias_variable('b_conv3', [512])\n",
296 | "\n",
297 | "conv1 = tf.nn.relu(conv2d(x_image, w_conv1, 2) + b_conv1)\n",
298 | "conv2 = tf.nn.relu(conv2d(conv1, w_conv2, 2) + b_conv2)\n",
299 | "conv3 = tf.nn.relu(conv2d(conv2, w_conv3, 2) + b_conv3)\n",
300 | "\n",
301 | "conv_size = conv3.get_shape()[1]\n",
302 | "conv_flat = tf.reshape(conv3, [-1, conv_size*conv_size, 512])\n",
303 | "conv_unstack = tf.unstack(conv_flat, axis = 1)\n",
304 | "\n",
305 | "#LSTM Variables\n",
306 | "Wf = weight_variable('Wf', [512 + lstm_size, lstm_size])\n",
307 | "Wi = weight_variable('Wi', [512 + lstm_size, lstm_size])\n",
308 | "Wc = weight_variable('Wc', [512 + lstm_size, lstm_size])\n",
309 | "Wo = weight_variable('Wo', [512 + lstm_size, lstm_size])\n",
310 | "\n",
311 | "bf = bias_variable('bf', [lstm_size])\n",
312 | "bi = bias_variable('bi', [lstm_size])\n",
313 | "bc = bias_variable('bc', [lstm_size])\n",
314 | "bo = bias_variable('bo', [lstm_size]) \n",
315 | "\n",
316 | "# Attention Variables\n",
317 | "Wa = weight_variable('Wa', [512, 1])\n",
318 | "Wh = weight_variable('Wh', [lstm_size, 1])\n",
319 | "\n",
320 | "rnn_batch_size = tf.shape(x_image)[0]\n",
321 | "\n",
322 | "# Initial lstm cell state and output \n",
323 | "rnn_state = tf.zeros([rnn_batch_size, lstm_size], tf.float32)\n",
324 | "rnn_out = tf.zeros([rnn_batch_size, lstm_size], tf.float32)\n",
325 | "\n",
326 | "#################################### Attention!!! ####################################\n",
327 | "for i in range(step_size):\n",
328 | " alpha, z = soft_attention(rnn_out, conv_unstack, Wa, Wh)\n",
329 | " rnn_state, rnn_out = LSTM_cell(rnn_state, rnn_out, z, Wf, Wi, Wc, Wo, bf, bi, bc, bo)\n",
330 | "\n",
331 | "######################################################################################\n",
332 | "\n",
333 | "# Densely connect layer variables \n",
334 | "w_fc1 = weight_variable('w_fc1', [lstm_size, num_label])\n",
335 | "b_fc1 = bias_variable('b_fc1', [num_label])\n",
336 | "\n",
337 | "output = tf.matmul(rnn_out, w_fc1)+b_fc1\n",
338 | "\n",
339 | "# Training \n",
340 | "Loss = tf.nn.sigmoid_cross_entropy_with_logits(labels = y_target, logits = output)\n",
341 | "Cost = tf.reduce_mean(Loss)\n",
342 | "optimizer = tf.train.AdamOptimizer(learning_rate = learning_rate, epsilon = epsilon).minimize(Cost)\n",
343 | "\n",
344 | "correct_prediction = tf.equal(tf.argmax(y_target,1), tf.argmax(output,1))\n",
345 | "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n"
346 | ]
347 | },
348 | {
349 | "cell_type": "markdown",
350 | "metadata": {},
351 | "source": [
352 | "## Create Session"
353 | ]
354 | },
355 | {
356 | "cell_type": "code",
357 | "execution_count": 8,
358 | "metadata": {
359 | "collapsed": true
360 | },
361 | "outputs": [],
362 | "source": [
363 | "# Create Session\n",
364 | "config = tf.ConfigProto()\n",
365 | "config.gpu_options.per_process_gpu_memory_fraction = gpu_fraction\n",
366 | "\n",
367 | "sess = tf.InteractiveSession(config=config)\n",
368 | "sess.run(tf.global_variables_initializer())"
369 | ]
370 | },
371 | {
372 | "cell_type": "markdown",
373 | "metadata": {},
374 | "source": [
375 | "## Save and Restore"
376 | ]
377 | },
378 | {
379 | "cell_type": "code",
380 | "execution_count": 9,
381 | "metadata": {},
382 | "outputs": [
383 | {
384 | "name": "stdout",
385 | "output_type": "stream",
386 | "text": [
387 | "INFO:tensorflow:Restoring parameters from saved_networks/soft1\n",
388 | "Successfully loaded: saved_networks/soft1\n"
389 | ]
390 | }
391 | ],
392 | "source": [
393 | "# Load the file if the saved file exists\n",
394 | "saver = tf.train.Saver()\n",
395 | "if Load_model == True:\n",
396 | " checkpoint = tf.train.get_checkpoint_state(\"saved_networks/\")\n",
397 | " if checkpoint and checkpoint.model_checkpoint_path:\n",
398 | " saver.restore(sess, checkpoint.model_checkpoint_path)\n",
399 | " print(\"Successfully loaded:\", checkpoint.model_checkpoint_path)\n",
400 | " else:\n",
401 | " print(\"Could not find old network weights\")"
402 | ]
403 | },
404 | {
405 | "cell_type": "markdown",
406 | "metadata": {},
407 | "source": [
408 | "## Training"
409 | ]
410 | },
411 | {
412 | "cell_type": "code",
413 | "execution_count": 10,
414 | "metadata": {
415 | "collapsed": true
416 | },
417 | "outputs": [],
418 | "source": [
419 | "# Training\n",
420 | "\n",
421 | "if Is_train == True:\n",
422 | " train_data_num = train_x.shape[0]\n",
423 | "\n",
424 | " for i in range(num_epoch):\n",
425 | " # Making batches\n",
426 | " random_idx = np.arange(train_data_num)\n",
427 | " np.random.shuffle(random_idx)\n",
428 | "\n",
429 | " batch_count = 1\n",
430 | " \n",
431 | " for j in range(0, train_data_num, batch_size):\n",
432 | " if j + batch_size < train_data_num:\n",
433 | " batch_index = [j, j + batch_size]\n",
434 | "\n",
435 | " batch_x_train = train_x[random_idx[batch_index[0]:batch_index[1]],:,:]\n",
436 | " batch_y_train = train_y[random_idx[batch_index[0]:batch_index[1]],:]\n",
437 | " else:\n",
438 | " batch_index = [j, j + train_data_num-1]\n",
439 | "\n",
440 | " batch_x_train = train_x[random_idx[batch_index[0]:batch_index[-1]],:,:]\n",
441 | " batch_y_train = train_y[random_idx[batch_index[0]:batch_index[-1]],:]\n",
442 | "\n",
443 | " # Make image as fractions for attention\n",
444 | " train_batch = np.reshape(batch_x_train, (batch_x_train.shape[0], img_size, img_size, 1))\n",
445 | " validation_batch = np.reshape(validation_x, (validation_x.shape[0], img_size, img_size, 1))\n",
446 | " \n",
447 | " # Training\n",
448 | " optimizer.run(feed_dict = {x_image: train_batch, y_target: batch_y_train})\n",
449 | " cost = sess.run(Cost, feed_dict = {x_image: train_batch, y_target: batch_y_train})\n",
450 | " acc = sess.run(accuracy, feed_dict = {x_image: train_batch, y_target: batch_y_train})\n",
451 | " val_acc = sess.run(accuracy, feed_dict = {x_image: validation_batch, y_target: validation_y})\n",
452 | "\n",
453 | " # Print Progress\n",
454 | " print(\"Epoch: \" + str(i+1) + ' / ' + \n",
455 | " \"Batch: \" + str(j) + '/' + str(train_data_num) + ' / ' + \n",
456 | " \"Cost: \" + str(cost) + ' / ' + \n",
457 | " \"Training Accuracy: \" + str(acc) + ' / ' + \n",
458 | " \"Validation Accuracy: \" + str(val_acc)) \n",
459 | "\n",
460 | " saver.save(sess, 'saved_networks/' + save_name)\n",
461 | " print('Model is saved!!!')"
462 | ]
463 | },
464 | {
465 | "cell_type": "markdown",
466 | "metadata": {},
467 | "source": [
468 | "## Testing"
469 | ]
470 | },
471 | {
472 | "cell_type": "code",
473 | "execution_count": 11,
474 | "metadata": {
475 | "scrolled": false
476 | },
477 | "outputs": [
478 | {
479 | "data": {
480 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDoWeEEIaDoWeEEIaDoWeEEIazv8HGuqvtGxkcHUAAAAASUVORK5CYII=\n",
481 | "text/plain": [
482 | ""
483 | ]
484 | },
485 | "metadata": {},
486 | "output_type": "display_data"
487 | },
488 | {
489 | "data": {
490 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
491 | "text/plain": [
492 | ""
493 | ]
494 | },
495 | "metadata": {},
496 | "output_type": "display_data"
497 | },
498 | {
499 | "data": {
500 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCSGk4VDoCSGk4VDoCSGk4VDoCSGk4fw/vmI62n26u4kAAAAASUVORK5CYII=\n",
501 | "text/plain": [
502 | ""
503 | ]
504 | },
505 | "metadata": {},
506 | "output_type": "display_data"
507 | },
508 | {
509 | "data": {
510 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n6IUQoub8P8/lh9XfSvz7AAAAAElFTkSuQmCC\n",
511 | "text/plain": [
512 | ""
513 | ]
514 | },
515 | "metadata": {},
516 | "output_type": "display_data"
517 | },
518 | {
519 | "data": {
520 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
521 | "text/plain": [
522 | ""
523 | ]
524 | },
525 | "metadata": {},
526 | "output_type": "display_data"
527 | },
528 | {
529 | "data": {
530 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBm6iAAAAAElFTkSuQmCC\n",
531 | "text/plain": [
532 | ""
533 | ]
534 | },
535 | "metadata": {},
536 | "output_type": "display_data"
537 | },
538 | {
539 | "data": {
540 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n9LSSAAAAAElFTkSuQmCC\n",
541 | "text/plain": [
542 | ""
543 | ]
544 | },
545 | "metadata": {},
546 | "output_type": "display_data"
547 | },
548 | {
549 | "data": {
550 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCGk4/w8VPXbHUg1bnwAAAABJRU5ErkJggg==\n",
551 | "text/plain": [
552 | ""
553 | ]
554 | },
555 | "metadata": {},
556 | "output_type": "display_data"
557 | },
558 | {
559 | "data": {
560 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nlOvwYHDwAAAAAElFTkSuQmCC\n",
561 | "text/plain": [
562 | ""
563 | ]
564 | },
565 | "metadata": {},
566 | "output_type": "display_data"
567 | },
568 | {
569 | "data": {
570 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADHCAYAAAAXg5iPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEJI4FHpCCEmc/wcojwB5W0FrtQAAAABJRU5ErkJggg==\n",
571 | "text/plain": [
572 | ""
573 | ]
574 | },
575 | "metadata": {},
576 | "output_type": "display_data"
577 | },
578 | {
579 | "name": "stdout",
580 | "output_type": "stream",
581 | "text": [
582 | "Sample Accuracy: 1.0\n"
583 | ]
584 | }
585 | ],
586 | "source": [
587 | "# Sampling test indexes\n",
588 | "idx = random.sample(range(test_x.shape[0]), num_test_sample)\n",
589 | "\n",
590 | "# Initialize fraction of test images and heatmap\n",
591 | "test_fraction = np.zeros([10, img_size, img_size, 1])\n",
592 | "heat_map = np.zeros([num_test_sample, img_size, img_size])\n",
593 | "\n",
594 | "num_correct = 0.\n",
595 | "\n",
596 | "# Test for Sampling data\n",
597 | "for idx_sample in range(num_test_sample):\n",
598 | " \n",
599 | " # Get alpha(weight of fractions) and output for sample test data\n",
600 | " test_x_reshape = np.reshape(test_x, ([test_x.shape[0],img_size,img_size,1]))\n",
601 | " test_x_in = test_x_reshape[idx[idx_sample],:,:,:]\n",
602 | " alpha_, output_ = sess.run([alpha, output],feed_dict = {x_image: [test_x_in], y_target: [test_y[idx[idx_sample],:]]})\n",
603 | " alpha_size = int(np.sqrt(alpha_.shape[1]))\n",
604 | " alpha_reshape = np.reshape(alpha_, (alpha_size, alpha_size))\n",
605 | " alpha_resize = skimage.transform.pyramid_expand(alpha_reshape, upscale = 16, sigma=20) \n",
606 | " \n",
607 | " # Make heatmap with alpha\n",
608 | "\n",
609 | " # Get labels for test samples\n",
610 | " y_test_pred = np.argmax(output_[:])\n",
611 | " y_test_true = np.argmax(test_y[idx[idx_sample], :])\n",
612 | " \n",
613 | " # Draw subplot for each sample \n",
614 | " f1, ax = plt.subplots(1,2)\n",
615 | " ax[0].imshow(alpha_resize, cmap='gray')\n",
616 | " ax[0].axis(\"off\")\n",
617 | " ax[0].set_title('Attention Heatmap')\n",
618 | " ax[1].imshow(test_x[idx[idx_sample],:,:], cmap='gray')\n",
619 | " ax[1].axis(\"off\")\n",
620 | " ax[1].set_title('Prediction: ' + str(y_test_pred) + ' / ' + 'Label: ' + str(y_test_true))\n",
621 | "\n",
622 | " # Count correct\n",
623 | " if y_test_pred == y_test_true:\n",
624 | " num_correct += 1.\n",
625 | "\n",
626 | "# Show results \n",
627 | "plt.show()\n",
628 | "print('Sample Accuracy: ' + str(num_correct / num_test_sample))"
629 | ]
630 | },
631 | {
632 | "cell_type": "code",
633 | "execution_count": null,
634 | "metadata": {
635 | "collapsed": true
636 | },
637 | "outputs": [],
638 | "source": []
639 | }
640 | ],
641 | "metadata": {
642 | "anaconda-cloud": {},
643 | "kernelspec": {
644 | "display_name": "Python 3",
645 | "language": "python",
646 | "name": "python3"
647 | },
648 | "language_info": {
649 | "codemirror_mode": {
650 | "name": "ipython",
651 | "version": 3
652 | },
653 | "file_extension": ".py",
654 | "mimetype": "text/x-python",
655 | "name": "python",
656 | "nbconvert_exporter": "python",
657 | "pygments_lexer": "ipython3",
658 | "version": "3.6.3"
659 | }
660 | },
661 | "nbformat": 4,
662 | "nbformat_minor": 1
663 | }
664 |
--------------------------------------------------------------------------------
/saved_networks/checkpoint:
--------------------------------------------------------------------------------
1 | model_checkpoint_path: "wormy"
2 | all_model_checkpoint_paths: "wormy"
3 |
--------------------------------------------------------------------------------