├── .gitignore ├── LICENSE ├── README.md ├── data_helper.py ├── img_cnn.py ├── pre_train.py └── train.py /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | env/ 12 | build/ 13 | develop-eggs/ 14 | dist/ 15 | downloads/ 16 | eggs/ 17 | .eggs/ 18 | lib/ 19 | lib64/ 20 | parts/ 21 | sdist/ 22 | var/ 23 | wheels/ 24 | *.egg-info/ 25 | .installed.cfg 26 | *.egg 27 | 28 | # PyInstaller 29 | # Usually these files are written by a python script from a template 30 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 31 | *.manifest 32 | *.spec 33 | 34 | # Installer logs 35 | pip-log.txt 36 | pip-delete-this-directory.txt 37 | 38 | # Unit test / coverage reports 39 | htmlcov/ 40 | .tox/ 41 | .coverage 42 | .coverage.* 43 | .cache 44 | nosetests.xml 45 | coverage.xml 46 | *.cover 47 | .hypothesis/ 48 | 49 | # Translations 50 | *.mo 51 | *.pot 52 | 53 | # Django stuff: 54 | *.log 55 | local_settings.py 56 | 57 | # Flask stuff: 58 | instance/ 59 | .webassets-cache 60 | 61 | # Scrapy stuff: 62 | .scrapy 63 | 64 | # Sphinx documentation 65 | docs/_build/ 66 | 67 | # PyBuilder 68 | target/ 69 | 70 | # Jupyter Notebook 71 | .ipynb_checkpoints 72 | 73 | # pyenv 74 | .python-version 75 | 76 | # celery beat schedule file 77 | celerybeat-schedule 78 | 79 | # SageMath parsed files 80 | *.sage.py 81 | 82 | # dotenv 83 | .env 84 | 85 | # virtualenv 86 | .venv 87 | venv/ 88 | ENV/ 89 | 90 | # Spyder project settings 91 | .spyderproject 92 | .spyproject 93 | 94 | # Rope project settings 95 | .ropeproject 96 | 97 | # mkdocs documentation 98 | /site 99 | 100 | # mypy 101 | .mypy_cache/ 102 | 103 | # custom 104 | inputs/ 105 | log/ 106 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright 2018 Lei Zhang 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # cnn-classification-dog-vs-cat 2 | 基于CNN的图像分类器，使用Kaggle的猫狗图片数据。 3 | 4 | ## 1 requirement 5 | - python3 6 | - numpy >= 1.14.2 7 | - keras >= 2.1.6 8 | - tensorflow >= 1.6.0 9 | - h5py >= 2.7.0 10 | - python-gflags >= 3.1.2 11 | - opencv-python >= 3.4.0.12 12 | 13 | ## 2 Description of files 14 | - inputs: 猫狗图片样本数据，[[下载地址]](https://www.kaggle.com/c/dogs-vs-cats/data)，使用keras库中的[ImageDataGenerator](https://keras.io/preprocessing/image/)类读取，需要将每个类的图片放在单独命名的文件夹中存放； 15 | - train.py: 自建的简单CNN，训练后测试集精度约83%； 16 | - pre_train.py: 利用已训练的常用网络(基于[ImageNet](http://www.image-net.org/)数据集训练)，进行迁移学习，测试集精度约95%以上； 17 | - data_helper.py: 数据读取和预处理模块； 18 | - img_cnn.py: 基于TensorFlow的自定义简单卷积神经网络。 19 | 20 | ## 3 Start training 21 | - ### 训练自定义的小型CNN 22 | ```shell 23 | python train.py 24 | ``` 25 | - ### 在VGG16的基础上进行迁移学习 26 | ```shell 27 | python pre_train.py 28 | ``` 29 | 30 | ## 4 Visualizing results in TensorBoard 31 | ```shell 32 | tensorboard --logdir /"PATH_OF_CODE"/log/"TIMESTAMP"/summaries/ 33 | ``` 34 | 35 | ## 5 References 36 | [1]. 猫狗图像数据来源： 37 | https://www.kaggle.com/c/dogs-vs-cats/data 38 | 39 | [2]. keras中载入已训练网络的方法： 40 | https://keras.io/applications/ 41 | 42 | [3]. keras中图像预处理的相关功能介绍： 43 | https://keras.io/preprocessing/image/ -------------------------------------------------------------------------------- /data_helper.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | 3 | import os 4 | import numpy as np 5 | import tensorflow as tf 6 | import cv2 7 | 8 | 9 | def get_filenames_and_labels(dir_path, folder_names=['cat', 'dog'], shuffle=True): 10 | img_path_list = [] 11 | label_list = [] 12 | img_count = 0 13 | for folder_name in folder_names: 14 | filenames = os.listdir(os.path.join(dir_path, folder_name)) 15 | for f in filenames: 16 | img_path_list.append(os.path.join(dir_path, folder_name, f)) 17 | label_list.append([0,1] if 'cat' in f else [1,0]) 18 | img_count += 1 19 | img_path_list = np.array(img_path_list) 20 | label_list = np.array(label_list) 21 | if shuffle == True: 22 | index = np.random.permutation(np.arange(0, img_count, 1)) 23 | img_path_list_shuffled = img_path_list[index] 24 | label_list_shuffled = label_list[index] 25 | else: 26 | img_path_list_shuffled = img_path_list 27 | label_list_shuffled = label_list 28 | return img_path_list_shuffled, label_list_shuffled 29 | 30 | def img_resize(img_path, img_height, img_width): 31 | img_src = cv2.imread(img_path) 32 | img_resized = cv2.resize(img_src, (img_height,img_width), interpolation=cv2.INTER_CUBIC) 33 | return img_resized 34 | 35 | def rgb2gray(img_rgb): 36 | img_gray = np.dot(img_rgb[...,:3], [0.299, 0.587, 0.114]) 37 | img_gray = img_gray / 255.0 38 | return img_gray.reshape(img_rgb.shape[0], img_rgb.shape[1], 1) 39 | 40 | def batch_iter(batch_size, num_epochs, img_path_list, label_list, 41 | img_height, img_width, shuffle=True): 42 | ''' 43 | Generates a batch iterator for a dataset. 44 | ''' 45 | img_path_list = np.array(img_path_list) 46 | label_list = np.array(label_list) 47 | data_size = len(label_list) 48 | num_batches_per_epoch = int((data_size-1)/batch_size)+1 49 | for epoch in range(num_epochs): 50 | if shuffle: 51 | shuffle_indices = np.random.permutation(np.arange(data_size)) 52 | img_path_list_shuffled = img_path_list[shuffle_indices] 53 | label_list_shuffled = label_list[shuffle_indices] 54 | else: 55 | img_path_list_shuffled = img_path_list 56 | label_list_shuffled = label_list 57 | for batch_num in range(num_batches_per_epoch): 58 | start_index = batch_num*batch_size 59 | end_index = min((batch_num+1)*batch_size, data_size) 60 | img_list_shuffled = [] 61 | for i in range(start_index, end_index): 62 | img_data = img_resize(img_path=img_path_list_shuffled[i], img_height=img_height, img_width=img_width) 63 | # img_data_min, img_data_max = np.min(img_data), np.max(img_data) 64 | # img_data = (img_data - img_data_min) / (img_data_max - img_data_min) 65 | img_data = rgb2gray(img_data) 66 | img_list_shuffled.append(img_data) 67 | img_list_shuffled = np.array(img_list_shuffled) 68 | yield img_list_shuffled, label_list_shuffled[start_index:end_index] 69 | 70 | def generate_arrays_from_file(batch_size, img_path_list, label_list, 71 | img_height, img_width, shuffle=True): 72 | ''' 73 | Generates a batch iterator for a dataset. 74 | ''' 75 | img_path_list = np.array(img_path_list) 76 | label_list = np.array(label_list) 77 | data_size = len(label_list) 78 | num_batches_per_epoch = int((data_size-1)/batch_size)+1 79 | while True: 80 | if shuffle: 81 | shuffle_indices = np.random.permutation(np.arange(data_size)) 82 | img_path_list_shuffled = img_path_list[shuffle_indices] 83 | label_list_shuffled = label_list[shuffle_indices] 84 | else: 85 | img_path_list_shuffled = img_path_list 86 | label_list_shuffled = label_list 87 | for batch_num in range(num_batches_per_epoch): 88 | start_index = batch_num*batch_size 89 | end_index = min((batch_num+1)*batch_size, data_size) 90 | img_list_shuffled = [] 91 | for i in range(start_index, end_index): 92 | img_data = img_resize(img_path=img_path_list_shuffled[i], img_height=img_height, img_width=img_width) 93 | # img_data_min, img_data_max = np.min(img_data), np.max(img_data) 94 | # img_data = (img_data - img_data_min) / (img_data_max - img_data_min) 95 | img_list_shuffled.append(img_data) 96 | img_list_shuffled = np.array(img_list_shuffled) 97 | yield ({'input_1': img_list_shuffled}, {'output': label_list_shuffled[start_index:end_index]}) 98 | -------------------------------------------------------------------------------- /img_cnn.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | 3 | import numpy as np 4 | import tensorflow as tf 5 | 6 | 7 | class ImgCNN(object): 8 | ''' 9 | A cnn for image classification. 10 | ''' 11 | def __init__(self, n_classes, img_height, img_width, img_channel, device_name='/cpu:0'): 12 | self.input_x = tf.placeholder(dtype=tf.float32, shape=[None, img_height, img_width, img_channel], name='input_x') 13 | self.input_y = tf.placeholder(dtype=tf.float32, shape=[None, n_classes], name='input_y') 14 | self.dropout_keep_prob = tf.placeholder(dtype=tf.float32, name='dropout_keep_prob') 15 | 16 | with tf.device(device_name): 17 | self.input_image = tf.reshape(self.input_x, [-1,img_height,img_width,img_channel]) 18 | with tf.name_scope('conv_layer_1'): 19 | filter_shape_1 = [5,5,img_channel,8] 20 | self.h_conv_1 = self.conv2d(x=self.input_x, W=self.w_variable(shape=filter_shape_1), stride=1, padding='SAME') 21 | self.h_conv_1 = tf.nn.relu(features=self.h_conv_1, name='relu_conv_1') 22 | with tf.name_scope('pooling_layer_1'): 23 | self.h_pool_1 = self.max_pool(x=self.h_conv_1, ksize=2, stride=2, padding='SAME') # shape: [112 * 112 * 8] 24 | 25 | with tf.name_scope('conv_layer_2'): 26 | filter_shape_2 = [3,3,8,16] 27 | self.h_conv_2 = self.conv2d(x=self.h_pool_1, W=self.w_variable(shape=filter_shape_2), stride=1, padding='SAME') 28 | self.h_conv_2 = tf.nn.relu(features=self.h_conv_2, name='relu_conv_2') 29 | with tf.name_scope('pooling_layer_2'): 30 | self.h_pool_2 = self.max_pool(x=self.h_conv_2, ksize=2, stride=2, padding='SAME') # shape: [56 * 56 * 16] 31 | 32 | with tf.name_scope('conv_layer_3'): 33 | filter_shape_3 = [3,3,16,32] 34 | self.h_conv_3 = self.conv2d(x=self.h_pool_2, W=self.w_variable(shape=filter_shape_3), stride=1, padding='SAME') 35 | self.h_conv_3 = tf.nn.relu(features=self.h_conv_3, name='relu_conv_3') 36 | with tf.name_scope('pooling_layer_3'): 37 | self.h_pool_3 = self.max_pool(x=self.h_conv_3, ksize=2, stride=2, padding='SAME') # shape: [28 * 28 * 32] 38 | 39 | num_total_unit = self.h_pool_3.get_shape()[1:4].num_elements() 40 | self.h_pool_3_flat = tf.reshape(self.h_pool_3, shape=[-1, num_total_unit]) 41 | 42 | with tf.name_scope('fc_layer_1'): 43 | self.h_fc_1 = self.fc_layer(self.h_pool_3_flat, num_total_unit, 128, activation_function=tf.nn.relu) 44 | 45 | with tf.name_scope('dropout'): 46 | self.h_drop = tf.nn.dropout(self.h_fc_1, keep_prob=self.dropout_keep_prob, name='h_drop') 47 | 48 | with tf.name_scope('fc_layer_2'): 49 | self.output = self.fc_layer(self.h_drop, 128, n_classes, activation_function=None) 50 | 51 | with tf.device('/cpu:0'): 52 | with tf.name_scope('prediction'): 53 | self.y_pred = tf.argmax(input=self.output, axis=1, name='y_pred') 54 | 55 | with tf.name_scope('loss'): 56 | self.loss = tf.reduce_mean( 57 | tf.nn.softmax_cross_entropy_with_logits(labels=self.input_y, logits=self.output), name='cross_entropy_loss') 58 | 59 | with tf.name_scope('accuracy'): 60 | correct_predictions = tf.equal(self.y_pred, tf.argmax(self.input_y, axis=1)) 61 | self.accuracy = tf.reduce_mean(tf.cast(correct_predictions, 'float'), name='accuracy') 62 | 63 | def w_variable(self, shape): 64 | return tf.Variable(initial_value=tf.truncated_normal(shape=shape, mean=0.0, stddev=0.1), dtype=tf.float32, name='W') 65 | 66 | def b_variable(self, shape): 67 | return tf.Variable(initial_value=tf.constant(value=0.1, shape=shape), dtype=tf.float32, name='b') 68 | 69 | def conv2d(self, x, W, stride, padding='SAME'): 70 | return tf.nn.conv2d(input=x, filter=W, strides=[1,stride,stride,1], padding=padding, name='conv') 71 | 72 | def max_pool(self, x, ksize, stride, padding='VALID'): 73 | return tf.nn.max_pool(value=x, ksize=[1,ksize,ksize,1], strides=[1,stride,stride,1], padding=padding, name='max-pool') 74 | 75 | def fc_layer(self, x, in_size, out_size, activation_function=None): 76 | w = self.w_variable(shape=[in_size, out_size]) 77 | b = self.b_variable(shape=[out_size]) 78 | z = tf.nn.xw_plus_b(x, w, b, name='Wx_plus_b') 79 | if activation_function is None: 80 | outputs = z 81 | else: 82 | outputs = activation_function(z) 83 | return outputs 84 | -------------------------------------------------------------------------------- /pre_train.py: -------------------------------------------------------------------------------- 1 | # coding:utf-8 2 | 3 | import sys 4 | import gflags 5 | import keras 6 | import matplotlib.pyplot as plt 7 | 8 | ### parameters ### 9 | # =============================================== 10 | FLAGS = gflags.FLAGS 11 | 12 | # data loading parameters 13 | gflags.DEFINE_string('train_data_dir', './inputs/train/', 14 | 'Directory of the training data.') 15 | gflags.DEFINE_string('dev_data_dir', './inputs/dev/', 16 | 'Directory of the dev data.') 17 | # gflags.DEFINE_float('dev_sample_percentage', 0.02, 'Percentage of the training data to user for validation (dev set).') 18 | 19 | # model parameters 20 | gflags.DEFINE_integer('img_height', 224, 21 | 'The height of the image for training (default: 227).') 22 | gflags.DEFINE_integer('img_width', 224, 23 | 'The width of the image for training (default: 227).') 24 | gflags.DEFINE_integer( 25 | 'img_channels', 3, 26 | 'The number of channels of the image for training (default: 3).') 27 | gflags.DEFINE_float('dropout_keep_prob', 0.7, 28 | 'Dropout keep probability (default: 0.7).') 29 | 30 | # training parameters 31 | gflags.DEFINE_float('learning_rate', 0.001, 'Learning rate for training.') 32 | gflags.DEFINE_integer('batch_size', 32, 'The batch size for each train step.') 33 | gflags.DEFINE_integer('num_epochs', 10, 34 | 'Number of training epochs (default: 200).') 35 | 36 | FLAGS(sys.argv) 37 | # show parameters 38 | print('\nPARAMETERS:') 39 | print('================================') 40 | for attr, value in FLAGS.flag_values_dict().items(): 41 | print('{0}: {1}'.format(attr.lower(), value)) 42 | print('================================\n\n') 43 | 44 | ### use the pre-trained model 45 | # create the base pre-trained model 46 | base_model = keras.applications.VGG16( 47 | weights='imagenet', 48 | include_top=False, 49 | input_shape=(FLAGS.img_height, FLAGS.img_width, FLAGS.img_channels)) 50 | 51 | # add a global spatial average pooling layer 52 | add_model = keras.Sequential(name='additional_layers') 53 | add_model.add(keras.layers.Flatten(input_shape=base_model.output_shape[1:])) 54 | add_model.add(keras.layers.Dense(128, activation='relu')) 55 | add_model.add(keras.layers.Dense(2, activation='softmax')) 56 | 57 | model = keras.models.Model( 58 | inputs=base_model.input, outputs=add_model(base_model.output)) 59 | 60 | # freeze all VGG16 layers 61 | for layer in model.layers[:-1]: 62 | layer.trainable = False 63 | 64 | model.compile( 65 | loss='categorical_crossentropy', 66 | optimizer=keras.optimizers.SGD(lr=FLAGS.learning_rate, momentum=0.9), 67 | metrics=['accuracy']) 68 | 69 | model.summary() 70 | 71 | train_datagen = keras.preprocessing.image.ImageDataGenerator( 72 | rescale=1. / 255, shear_range=0.2, zoom_range=0.2, horizontal_flip=True) 73 | validation_datagen = keras.preprocessing.image.ImageDataGenerator(rescale=1./255) 74 | train_generator = train_datagen.flow_from_directory( 75 | directory=FLAGS.train_data_dir, 76 | target_size=(FLAGS.img_height, FLAGS.img_width), 77 | batch_size=FLAGS.batch_size, 78 | class_mode='categorical', 79 | seed=272) 80 | validation_generator = validation_datagen.flow_from_directory( 81 | directory=FLAGS.dev_data_dir, 82 | target_size=(FLAGS.img_height, FLAGS.img_width), 83 | batch_size=FLAGS.batch_size, 84 | class_mode='categorical') 85 | 86 | # train the model on the new data for a few epochs 87 | history = model.fit_generator( 88 | # data_helper.generate_arrays_from_file( 89 | # batch_size=FLAGS.batch_size, img_path_list=x_path_train, label_list=y_train, img_height=224, img_width=224), 90 | train_generator, 91 | steps_per_epoch=train_generator.n // FLAGS.batch_size, 92 | epochs=FLAGS.num_epochs, 93 | validation_data=validation_generator, 94 | verbose=1, 95 | callbacks=[ 96 | keras.callbacks.ModelCheckpoint( 97 | './log/VGG16-transfer-learning.model', 98 | monitor='val_loss', 99 | save_best_only=True, 100 | verbose=1) 101 | ]) 102 | 103 | # summarize history for accuracy 104 | plt.plot(history.history['acc']) 105 | plt.plot(history.history['val_acc']) 106 | plt.title('model accuracy') 107 | plt.ylabel('accuracy') 108 | plt.xlabel('epoch') 109 | plt.legend(['train', 'validation'], loc='upper left') 110 | plt.show() 111 | 112 | # summarize history for loss 113 | plt.plot(history.history['loss']) 114 | plt.plot(history.history['val_loss']) 115 | plt.title('model loss') 116 | plt.ylabel('loss') 117 | plt.xlabel('epoch') 118 | plt.legend(['train', 'validation'], loc='upper left') 119 | plt.show() 120 | 121 | print("Training loss: {:.2f} / Validation loss: {:.2f}".format( 122 | history.history['loss'][-1], history.history['val_loss'][-1])) 123 | print("Training accuracy: {:.2f}% / Validation accuracy: {:.2f}%".format( 124 | 100 * history.history['acc'][-1], 100 * history.history['val_acc'][-1])) 125 | -------------------------------------------------------------------------------- /train.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | 3 | import sys 4 | import os 5 | import time 6 | import datetime 7 | import gflags 8 | import numpy as np 9 | import tensorflow as tf 10 | import data_helper 11 | from img_cnn import ImgCNN 12 | 13 | 14 | ### parameters ### 15 | # =============================================== 16 | FLAGS = gflags.FLAGS 17 | 18 | # data loading parameters 19 | gflags.DEFINE_string('train_data_dir', './inputs/train/', 'Directory of the training data.') 20 | gflags.DEFINE_float('dev_sample_percentage', 0.01, 'Percentage of the training data to user for validation (dev set).') 21 | 22 | # model parameters 23 | gflags.DEFINE_integer('img_height', 224, 'The height of the image for training (default: 227).') 24 | gflags.DEFINE_integer('img_width', 224, 'The width of the image for training (default: 227).') 25 | gflags.DEFINE_integer('img_channels', 1, 'The number of channels of the image for training (default: 3).') 26 | gflags.DEFINE_float('dropout_keep_prob', 0.7, 'Dropout keep probability (default: 0.7).') 27 | 28 | # training parameters 29 | gflags.DEFINE_float('learning_rate', 0.001, 'Learning rate for training.') 30 | gflags.DEFINE_integer('batch_size', 32, 'The batch size for each train step.') 31 | gflags.DEFINE_integer('num_epochs', 200, 'Number of training epochs (default: 200).') 32 | gflags.DEFINE_integer('evaluate_every', 100, 'Evaluate model on dev set after this many of steps (default: 100).') 33 | gflags.DEFINE_integer('checkpoint_every', 100, 'Save model after this many steps (default: 100).') 34 | gflags.DEFINE_integer('num_checkpoints', 5, 'Number of checkpoints to store (default: 5).') 35 | 36 | # device parameters 37 | gflags.DEFINE_string('device_name', '/cpu:0', 'Device name for training.') 38 | gflags.DEFINE_bool('allow_soft_placement', True, 'Allow device soft device placement.') 39 | gflags.DEFINE_bool('log_device_placement', False, 'Log placement of ops on devices.') 40 | 41 | FLAGS(sys.argv) 42 | # show parameters 43 | print('\nPARAMETERS:') 44 | print('================================') 45 | for attr, value in FLAGS.flag_values_dict().items(): 46 | print('{0}: {1}'.format(attr.lower(), value)) 47 | print('================================\n\n') 48 | 49 | 50 | ### data preparation ### 51 | # =============================================== 52 | 53 | # load data 54 | print('Loading data...\n') 55 | x_path, y = data_helper.get_filenames_and_labels(FLAGS.train_data_dir) 56 | 57 | # split train/dev set 58 | split_index = -int(float(len(y)) * FLAGS.dev_sample_percentage) 59 | x_path_train, x_path_dev = x_path[:split_index], x_path[split_index:] 60 | y_train, y_dev = y[:split_index], y[split_index:] 61 | 62 | del x_path, y 63 | 64 | x_dev = [] 65 | for i in range(len(x_path_dev)): 66 | img_data = data_helper.img_resize(img_path=x_path_dev[i], img_height=FLAGS.img_height, img_width=FLAGS.img_width) 67 | #img_data_min, img_data_max = np.min(img_data), np.max(img_data) 68 | #img_data = (img_data - img_data_min) / (img_data_max - img_data_min) 69 | img_data = data_helper.rgb2gray(img_data) 70 | x_dev.append(img_data) 71 | x_dev = np.array(x_dev) 72 | y_dev = np.array(y_dev) 73 | 74 | 75 | input('press enter to start training...\n\n') 76 | ### training 77 | # =============================================== 78 | print('start training...\n') 79 | with tf.Graph().as_default(): 80 | session_conf = tf.ConfigProto( 81 | allow_soft_placement=FLAGS.allow_soft_placement, 82 | log_device_placement=FLAGS.log_device_placement) 83 | sess = tf.Session(config=session_conf) 84 | with sess.as_default(): 85 | cnn = ImgCNN( 86 | n_classes=y_train.shape[1], 87 | img_height=FLAGS.img_height, 88 | img_width=FLAGS.img_width, 89 | img_channel=FLAGS.img_channels, 90 | device_name=FLAGS.device_name 91 | ) 92 | 93 | # define training procedure 94 | global_step = tf.Variable(0, trainable=False, name='global_step') 95 | optimizer = tf.train.AdamOptimizer(learning_rate=FLAGS.learning_rate) 96 | grads_and_vars = optimizer.compute_gradients(cnn.loss) 97 | train_op = optimizer.apply_gradients(grads_and_vars, global_step=global_step) 98 | 99 | # output directory for models and summaries 100 | timestamp = str(int(time.time())) 101 | out_dir = os.path.abspath(os.path.join(os.curdir, 'log', timestamp)) 102 | print('Writing log to {}\n'.format(out_dir)) 103 | 104 | # summary the input images 105 | tf.summary.image('input_image', cnn.input_image, max_outputs=FLAGS.batch_size) 106 | 107 | # summary all the trainable variables 108 | for var in tf.trainable_variables(): 109 | tf.summary.histogram(name=var.name, values=var) 110 | 111 | # summary loss and accuracy 112 | loss_summary = tf.summary.scalar('loss', cnn.loss) 113 | acc_summary = tf.summary.scalar('accuracy', cnn.accuracy) 114 | 115 | # train summaries 116 | # train_summary_op = tf.summary.merge([loss_summary, acc_summary]) 117 | train_summary_op = tf.summary.merge_all() 118 | train_summary_dir = os.path.join(out_dir, 'summaries', 'train') 119 | train_summary_writer = tf.summary.FileWriter(train_summary_dir, tf.get_default_graph()) 120 | 121 | # test summaries 122 | # dev_summary_op = tf.summary.merge([loss_summary, acc_summary]) 123 | dev_summary_op = tf.summary.merge_all() 124 | dev_summary_dir = os.path.join(out_dir, 'summaries', 'dev') 125 | dev_summary_writer = tf.summary.FileWriter(dev_summary_dir, tf.get_default_graph()) 126 | 127 | # checkpointing, tensorflow assumes this directory already existed, so we need to create it 128 | checkpoint_dir = os.path.join(out_dir, 'checkpoints') 129 | checkpoint_prefix = os.path.join(checkpoint_dir, 'model') 130 | if not os.path.exists(checkpoint_dir): 131 | os.makedirs(checkpoint_dir) 132 | saver = tf.train.Saver(tf.global_variables(), max_to_keep=FLAGS.num_checkpoints) 133 | 134 | def train_step(x_batch, y_batch, writer=None): 135 | ''' 136 | A single training step. 137 | ''' 138 | feed_dict = { 139 | cnn.input_x: x_batch, 140 | cnn.input_y: y_batch, 141 | cnn.dropout_keep_prob: FLAGS.dropout_keep_prob 142 | } 143 | _, step, summaries, loss, accuracy = sess.run( 144 | [train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy], 145 | feed_dict) 146 | timestr = datetime.datetime.now().isoformat() 147 | print('{}: step {}, loss {:g}, acc {:g}'.format(timestr, step, loss, accuracy)) 148 | if writer: 149 | writer.add_summary(summaries, step) 150 | 151 | def dev_step(x_batch, y_batch, writer=None): 152 | ''' 153 | Evaluate the model on test set. 154 | ''' 155 | feed_dict = { 156 | cnn.input_x: x_batch, 157 | cnn.input_y: y_batch, 158 | cnn.dropout_keep_prob: 1.0 159 | } 160 | step, summaries, loss, accuracy = sess.run( 161 | [global_step, dev_summary_op, cnn.loss, cnn.accuracy], 162 | feed_dict) 163 | timestr = datetime.datetime.now().isoformat() 164 | print('{}: step {}, loss {:g}, acc {:g}'.format(timestr, step, loss, accuracy)) 165 | if writer: 166 | writer.add_summary(summaries, step) 167 | 168 | ### training loop 169 | # train loop, for each batch 170 | sess.run(tf.global_variables_initializer()) 171 | batches = data_helper.batch_iter(batch_size=FLAGS.batch_size, num_epochs=FLAGS.num_epochs, img_path_list=x_path_train, label_list=y_train, 172 | img_height=FLAGS.img_height, img_width=FLAGS.img_width) 173 | for x_batch, y_batch in batches: 174 | train_step(x_batch, y_batch, writer=train_summary_writer) 175 | current_step = tf.train.global_step(sess, global_step) 176 | if current_step % FLAGS.evaluate_every == 0: 177 | print('\nEvaluation on dev set:') 178 | dev_step(x_dev, y_dev, writer=dev_summary_writer) 179 | print('') 180 | if current_step % FLAGS.checkpoint_every == 0: 181 | path = saver.save(sess=sess, save_path=checkpoint_prefix, global_step=global_step) 182 | print('\nSaved model checkpoint to {}\n'.format(path)) 183 | 184 | # end 185 | print('\n--- Done! ---\n') 186 | --------------------------------------------------------------------------------