├── LICENSE ├── README.md ├── get_predictions.ipynb ├── results ├── 1.png ├── 2.png ├── 3.png └── 4.png ├── test_anchors.ipynb ├── train.ipynb └── visualize.ipynb /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 [yyyy] [name of copyright owner] 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 | # ESRI Object Detection Challenge 2019 2 | I used [keras-retinanet](https://github.com/fizyr/keras-retinanet) to fine tune a pretrained RetinaNet model on this dataset. It achieves an AP of 83% on test set with IoU Threshold of 0.3. [Link to Dataset](https://www.kaggle.com/kbhartiya83/swimming-pool-and-car-detection) 3 | 4 | ## Requirements 5 | - Everything required by [keras-retinanet](https://github.com/fizyr/keras-retinanet) 6 | - TQDM 7 | - imgaug 8 | - sklearn 9 | 10 | ## Results 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | -------------------------------------------------------------------------------- /get_predictions.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "markdown", 5 | "metadata": {}, 6 | "source": [ 7 | "# Generate Predictions from Trained Model" 8 | ] 9 | }, 10 | { 11 | "cell_type": "code", 12 | "execution_count": null, 13 | "metadata": {}, 14 | "outputs": [], 15 | "source": [ 16 | "import os\n", 17 | "from os import listdir, walk\n", 18 | "from os.path import join\n", 19 | "from keras_retinanet.bin.train import create_models\n", 20 | "from keras_retinanet.models import backbone,convert_model\n", 21 | "from keras_retinanet.utils.config import read_config_file,parse_anchor_parameters\n", 22 | "import numpy as np\n", 23 | "import keras\n", 24 | "import math\n", 25 | "import tensorflow as tf\n", 26 | "tf.set_random_seed(31)\n", 27 | "np.random.seed(17)" 28 | ] 29 | }, 30 | { 31 | "cell_type": "code", 32 | "execution_count": null, 33 | "metadata": {}, 34 | "outputs": [], 35 | "source": [ 36 | "model, training_model, prediction_model = create_models(\n", 37 | " backbone_retinanet=backbone('resnet50').retinanet,\n", 38 | " num_classes=2,\n", 39 | " weights=None,\n", 40 | " multi_gpu=False,\n", 41 | " freeze_backbone=False,\n", 42 | " lr=1e-3,\n", 43 | " config=read_config_file('config.ini')\n", 44 | " )" 45 | ] 46 | }, 47 | { 48 | "cell_type": "markdown", 49 | "metadata": {}, 50 | "source": [ 51 | "# Load and Convert Model\n", 52 | "\n", 53 | "Training model needs to be converted to inference model" 54 | ] 55 | }, 56 | { 57 | "cell_type": "code", 58 | "execution_count": null, 59 | "metadata": {}, 60 | "outputs": [], 61 | "source": [ 62 | "training_model.load_weights('saved/resnet50_csv_70.h5')" 63 | ] 64 | }, 65 | { 66 | "cell_type": "code", 67 | "execution_count": null, 68 | "metadata": {}, 69 | "outputs": [], 70 | "source": [ 71 | "infer_model = convert_model(training_model,anchor_params=parse_anchor_parameters(read_config_file('config.ini')))" 72 | ] 73 | }, 74 | { 75 | "cell_type": "markdown", 76 | "metadata": {}, 77 | "source": [ 78 | "# Create Generator to load images" 79 | ] 80 | }, 81 | { 82 | "cell_type": "code", 83 | "execution_count": null, 84 | "metadata": {}, 85 | "outputs": [], 86 | "source": [ 87 | "from keras_retinanet.utils.image import read_image_bgr, preprocess_image, resize_image\n", 88 | "from tqdm import tqdm\n", 89 | "\n", 90 | "def test_gen(image_ids, bs = 32, size=672,test = True):\n", 91 | " imgs = []\n", 92 | " scale = None\n", 93 | " idx = 0\n", 94 | " if test:\n", 95 | " path = 'test_data_images/images/'\n", 96 | " else:\n", 97 | " path = 'training_data/images/'\n", 98 | " \n", 99 | " while idx < len(image_ids):\n", 100 | " if len(imgs) < bs:\n", 101 | " imgs.append(resize_image(preprocess_image(read_image_bgr(path + image_ids[idx] + '.jpg')),min_side=size,max_side=size)[0]) \n", 102 | " if scale is None:\n", 103 | " scale = resize_image(preprocess_image(read_image_bgr(path + image_ids[idx] + '.jpg')),min_side=size,max_side=size)[1]\n", 104 | " idx += 1\n", 105 | " else:\n", 106 | " yield np.array(imgs),scale\n", 107 | " imgs = []\n", 108 | " \n", 109 | " \n", 110 | " if len(imgs) > 0:\n", 111 | " yield np.array(imgs),scale " 112 | ] 113 | }, 114 | { 115 | "cell_type": "markdown", 116 | "metadata": {}, 117 | "source": [ 118 | "# Save Predictions" 119 | ] 120 | }, 121 | { 122 | "cell_type": "code", 123 | "execution_count": null, 124 | "metadata": {}, 125 | "outputs": [], 126 | "source": [ 127 | "_,_,image_ids = next(walk('test_data_images/images'))\n", 128 | "image_ids = [i[:-4] for i in image_ids]\n", 129 | "image_ids = sorted(image_ids)\n", 130 | "\n", 131 | "os.system('mkdir predictions')\n", 132 | "\n", 133 | "iter_num = 0\n", 134 | "test_bs = 8\n", 135 | "\n", 136 | "\n", 137 | "for imgs,scale in tqdm(test_gen(image_ids,bs=test_bs),total=math.ceil(len(image_ids)/test_bs)):\n", 138 | " boxes, scores, labels = infer_model.predict_on_batch(imgs)\n", 139 | " boxes /= scale\n", 140 | " for img_num in range(len(imgs)):\n", 141 | " with open('predictions/' + image_ids[(iter_num*test_bs) + img_num] + '.txt', 'w') as f:\n", 142 | " for box, score, label in zip(boxes[img_num], scores[img_num], labels[img_num]):\n", 143 | " # scores are sorted so we can break\n", 144 | " if score < 0:\n", 145 | " break\n", 146 | " f.write(f'{label + 1} {score} {int((box[1]))} {int((box[0]))} {int((box[3]))} {int((box[2]))} \\n')\n", 147 | " iter_num += 1" 148 | ] 149 | }, 150 | { 151 | "cell_type": "code", 152 | "execution_count": null, 153 | "metadata": {}, 154 | "outputs": [], 155 | "source": [ 156 | "os.system('zip -r predictions.zip predictions/')" 157 | ] 158 | }, 159 | { 160 | "cell_type": "code", 161 | "execution_count": null, 162 | "metadata": {}, 163 | "outputs": [], 164 | "source": [] 165 | } 166 | ], 167 | "metadata": { 168 | "kernelspec": { 169 | "display_name": "Python 3", 170 | "language": "python", 171 | "name": "python3" 172 | }, 173 | "language_info": { 174 | "codemirror_mode": { 175 | "name": "ipython", 176 | "version": 3 177 | }, 178 | "file_extension": ".py", 179 | "mimetype": "text/x-python", 180 | "name": "python", 181 | "nbconvert_exporter": "python", 182 | "pygments_lexer": "ipython3", 183 | "version": "3.6.7" 184 | } 185 | }, 186 | "nbformat": 4, 187 | "nbformat_minor": 2 188 | } 189 | -------------------------------------------------------------------------------- /results/1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/kunwar31/ESRI_Object_Detection/c78451d94ee1a5a2df344f7ef9849ad140a5b379/results/1.png -------------------------------------------------------------------------------- /results/2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/kunwar31/ESRI_Object_Detection/c78451d94ee1a5a2df344f7ef9849ad140a5b379/results/2.png -------------------------------------------------------------------------------- /results/3.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/kunwar31/ESRI_Object_Detection/c78451d94ee1a5a2df344f7ef9849ad140a5b379/results/3.png -------------------------------------------------------------------------------- /results/4.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/kunwar31/ESRI_Object_Detection/c78451d94ee1a5a2df344f7ef9849ad140a5b379/results/4.png -------------------------------------------------------------------------------- /test_anchors.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "markdown", 5 | "metadata": {}, 6 | "source": [ 7 | "# Visualize Anchor Boxes\n", 8 | "\n", 9 | "!OpenCV must be installed to run this notebook!" 10 | ] 11 | }, 12 | { 13 | "cell_type": "code", 14 | "execution_count": null, 15 | "metadata": {}, 16 | "outputs": [], 17 | "source": [ 18 | "from keras_retinanet.bin.train import create_generators\n", 19 | "from keras_retinanet.models import backbone\n", 20 | "from keras_retinanet.utils.config import read_config_file,parse_anchor_parameters\n", 21 | "from keras_retinanet.bin.debug import run" 22 | ] 23 | }, 24 | { 25 | "cell_type": "code", 26 | "execution_count": null, 27 | "metadata": {}, 28 | "outputs": [], 29 | "source": [ 30 | "b = backbone('resnet50')\n", 31 | "\n", 32 | "class args:\n", 33 | " batch_size = 4\n", 34 | " config = read_config_file('config.ini')\n", 35 | " random_transform = True\n", 36 | " annotations = 'annotations.csv'\n", 37 | " val_annotations = 'val_annotations.csv'\n", 38 | " classes = 'classes.csv'\n", 39 | " image_min_side = 672\n", 40 | " image_max_side = 672\n", 41 | " dataset_type = 'csv'\n", 42 | " tensorboard_dir = ''\n", 43 | " evaluation = True\n", 44 | " snapshots = True\n", 45 | " snapshot_path = \"saved/\"\n", 46 | " backbone = 'resnet50'\n", 47 | " weighted_average = True\n", 48 | " resize=True\n", 49 | " anchors=True\n", 50 | " \n", 51 | "train_gen,valid_gen = create_generators(args,b.preprocess_image)" 52 | ] 53 | }, 54 | { 55 | "cell_type": "code", 56 | "execution_count": null, 57 | "metadata": {}, 58 | "outputs": [], 59 | "source": [ 60 | "while run(valid_gen,args,parse_anchor_parameters(args.config)):\n", 61 | " pass" 62 | ] 63 | }, 64 | { 65 | "cell_type": "code", 66 | "execution_count": null, 67 | "metadata": {}, 68 | "outputs": [], 69 | "source": [] 70 | } 71 | ], 72 | "metadata": { 73 | "kernelspec": { 74 | "display_name": "Python 3", 75 | "language": "python", 76 | "name": "python3" 77 | }, 78 | "language_info": { 79 | "codemirror_mode": { 80 | "name": "ipython", 81 | "version": 3 82 | }, 83 | "file_extension": ".py", 84 | "mimetype": "text/x-python", 85 | "name": "python", 86 | "nbconvert_exporter": "python", 87 | "pygments_lexer": "ipython3", 88 | "version": "3.6.7" 89 | } 90 | }, 91 | "nbformat": 4, 92 | "nbformat_minor": 2 93 | } 94 | -------------------------------------------------------------------------------- /train.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "markdown", 5 | "metadata": { 6 | "_uuid": "1fd9570466736b09c16d4ed020cb81aef494753d" 7 | }, 8 | "source": [ 9 | "# **ESRI DATA SCIENCE CHALLENGE 2019**\n", 10 | "\n", 11 | "\n", 12 | "In this challenge, we have been given 224x224 images containing cars and swimming pools labelled in PASCAL VOC Format. As we'll be using RetinaNet for this challenge, we'll make use of keras-retinanet package from github. \n", 13 | "* [Focal Loss for Dense Object Detection](https://arxiv.org/abs/1708.02002) - Research paper describing RetinaNet and Focal Loss which it uses\n", 14 | "* [keras-retinanet](https://github.com/fizyr/keras-retinanet) - Keras RetinaNet implimentation" 15 | ] 16 | }, 17 | { 18 | "cell_type": "code", 19 | "execution_count": null, 20 | "metadata": { 21 | "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19", 22 | "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5" 23 | }, 24 | "outputs": [], 25 | "source": [ 26 | "!git clone https://github.com/fizyr/keras-retinanet" 27 | ] 28 | }, 29 | { 30 | "cell_type": "code", 31 | "execution_count": null, 32 | "metadata": { 33 | "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0", 34 | "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a" 35 | }, 36 | "outputs": [], 37 | "source": [ 38 | "!pip3 install keras-retinanet/ --user" 39 | ] 40 | }, 41 | { 42 | "cell_type": "code", 43 | "execution_count": null, 44 | "metadata": { 45 | "_uuid": "fb2593dea04ab0e20ef4ee521dd8cba7f3bd987b" 46 | }, 47 | "outputs": [], 48 | "source": [ 49 | "!rm keras-retinanet/ -R" 50 | ] 51 | }, 52 | { 53 | "cell_type": "code", 54 | "execution_count": null, 55 | "metadata": { 56 | "_uuid": "907236db932c74a1170900b8e6d9cc60bd086206" 57 | }, 58 | "outputs": [], 59 | "source": [ 60 | "import xml.etree.ElementTree as ET\n", 61 | "import os\n", 62 | "import numpy as np\n", 63 | "import keras\n", 64 | "import math\n", 65 | "import tensorflow as tf\n", 66 | "import cv2\n", 67 | "from os import listdir, walk\n", 68 | "from os.path import join\n", 69 | "from keras_retinanet.bin.train import create_generators,create_models,create_callbacks\n", 70 | "from keras_retinanet.models import backbone,load_model,convert_model\n", 71 | "from keras_retinanet.utils.config import read_config_file,parse_anchor_parameters\n", 72 | "from keras_retinanet.utils.visualization import draw_boxes\n", 73 | "from sklearn.model_selection import train_test_split\n", 74 | "from imgaug import augmenters as iaa\n", 75 | "\n", 76 | "tf.set_random_seed(31) # SEEDS MAKE RESULTS MORE REPRODUCABLE\n", 77 | "np.random.seed(17)" 78 | ] 79 | }, 80 | { 81 | "cell_type": "code", 82 | "execution_count": null, 83 | "metadata": { 84 | "_uuid": "faef2a3eac3226eb2e8211a9c81f89b9a6a811b3" 85 | }, 86 | "outputs": [], 87 | "source": [ 88 | "classes = ['1','2']" 89 | ] 90 | }, 91 | { 92 | "cell_type": "markdown", 93 | "metadata": { 94 | "_uuid": "56a388a1b5fc53f9ffd6c207a87bae5b8ffbc543" 95 | }, 96 | "source": [ 97 | "# Load and Convert Annotations\n", 98 | "\n", 99 | "Here we load annotations given in PASCAL VOC Format and save them in CSV Format as required by keras-retinanet package" 100 | ] 101 | }, 102 | { 103 | "cell_type": "code", 104 | "execution_count": null, 105 | "metadata": { 106 | "_uuid": "7673f02c16e5bb5e3734f23d2a533326b9fba562" 107 | }, 108 | "outputs": [], 109 | "source": [ 110 | "def convert_annotation(image_id,filename):\n", 111 | " in_file = open('training_data/labels/%s.xml'%(image_id))\n", 112 | " out_file = open(filename, 'a')\n", 113 | " tree=ET.parse(in_file)\n", 114 | " root = tree.getroot()\n", 115 | " \n", 116 | " if root.iter('object') is not None:\n", 117 | " for obj in root.iter('object'):\n", 118 | " cls = obj.find('name').text\n", 119 | " if cls not in classes:\n", 120 | " continue\n", 121 | " cls_id = classes.index(cls)\n", 122 | " \n", 123 | " xmlbox = obj.find('bndbox')\n", 124 | " x1 = math.ceil(float(xmlbox.find('xmin').text))\n", 125 | " y1 = math.ceil(float(xmlbox.find('ymin').text))\n", 126 | " x2 = math.ceil(float(xmlbox.find('xmax').text))\n", 127 | " y2 = math.ceil(float(xmlbox.find('ymax').text))\n", 128 | " if x1 == x2 or y1 == y2:\n", 129 | " continue\n", 130 | " \n", 131 | " out_file.write(f'training_data/images/{image_id}.jpg,{x1},{y1},{x2},{y2},{cls}\\n')\n", 132 | " else:\n", 133 | " out_file.write(f'training_data/images/{image_id}.jpg,,,,,\\n')" 134 | ] 135 | }, 136 | { 137 | "cell_type": "markdown", 138 | "metadata": { 139 | "_uuid": "04461e1dd8304d4224c0ce6fe9a1da1040b8aa9d" 140 | }, 141 | "source": [ 142 | "# Training and Validation split\n", 143 | "\n", 144 | "Normally we would have 10-30% of our images in validation set but as we want best possible score we'll use all our images to train, as we have quite few training images already. " 145 | ] 146 | }, 147 | { 148 | "cell_type": "code", 149 | "execution_count": null, 150 | "metadata": { 151 | "_uuid": "32f27dd808794bf8659b9e668b0e490bc6a104b7" 152 | }, 153 | "outputs": [], 154 | "source": [ 155 | "_,_,image_ids = next(walk('training_data/images'))\n", 156 | "image_ids = [i[:-4] for i in image_ids]\n", 157 | "open('annotations.csv','w')\n", 158 | "open('val_annotations.csv','w')\n", 159 | "\n", 160 | "train_ids,val_ids = train_test_split(image_ids,random_state=31,test_size=0)\n", 161 | "\n", 162 | "for image_id in train_ids:\n", 163 | " convert_annotation(image_id,'annotations.csv')\n", 164 | " \n", 165 | "for image_id in val_ids:\n", 166 | " convert_annotation(image_id,'val_annotations.csv')\n", 167 | " \n", 168 | "print(len(train_ids),len(val_ids))" 169 | ] 170 | }, 171 | { 172 | "cell_type": "code", 173 | "execution_count": null, 174 | "metadata": { 175 | "_uuid": "93124e395a1fdc94926afe1b71179770c84d587b" 176 | }, 177 | "outputs": [], 178 | "source": [ 179 | "with open('classes.csv','w') as f:\n", 180 | " f.write('1,0\\n2,1\\n')" 181 | ] 182 | }, 183 | { 184 | "cell_type": "markdown", 185 | "metadata": { 186 | "_uuid": "add87f5c791fbe8da4793a384e983235587d1626" 187 | }, 188 | "source": [ 189 | "# Anchor Parameters\n", 190 | "\n", 191 | "1. Anchor parameters are used to decide how anchor boxes will be generated for the model.\n", 192 | "1. As we're dealing mostly small boxes with can be highly elongated, we'll change ratios and scales to fit our needs.\n", 193 | "1. test_anchors.ipynb is used to visualize anchors on ground truth boxes" 194 | ] 195 | }, 196 | { 197 | "cell_type": "code", 198 | "execution_count": null, 199 | "metadata": { 200 | "_uuid": "17398f319e28dd6e38948e2e09dc87481e1966bc" 201 | }, 202 | "outputs": [], 203 | "source": [ 204 | "with open('config.ini','w') as f:\n", 205 | " f.write('[anchor_parameters]\\nsizes = 32 64 128 256 512\\nstrides = 8 16 32 64 128\\nratios = 0.25 0.5 0.75 1 1.5 2 4 6 8 10\\nscales = 0.5 1 2\\n')" 206 | ] 207 | }, 208 | { 209 | "cell_type": "markdown", 210 | "metadata": { 211 | "_uuid": "867d2c56814d761227270459f837b787e616f4c5" 212 | }, 213 | "source": [ 214 | "# Some Hyperparameters\n", 215 | "\n", 216 | "We will rescale our images to 672x672 for better precision" 217 | ] 218 | }, 219 | { 220 | "cell_type": "code", 221 | "execution_count": null, 222 | "metadata": { 223 | "_uuid": "ab8e28dc982006ee8f8800a9021433f0d716fc48" 224 | }, 225 | "outputs": [], 226 | "source": [ 227 | "b = backbone('resnet50')\n", 228 | "\n", 229 | "class args:\n", 230 | " batch_size = 64\n", 231 | " config = read_config_file('config.ini')\n", 232 | " random_transform = True # Image augmentation\n", 233 | " annotations = 'annotations.csv'\n", 234 | " val_annotations = 'val_annotations.csv'\n", 235 | " classes = 'classes.csv'\n", 236 | " image_min_side = 672\n", 237 | " image_max_side = 672\n", 238 | " dataset_type = 'csv'\n", 239 | " tensorboard_dir = ''\n", 240 | " evaluation = False\n", 241 | " snapshots = True\n", 242 | " snapshot_path = \"saved/\"\n", 243 | " backbone = 'resnet50'\n", 244 | " epochs = 70\n", 245 | " steps = len(train_ids)//(batch_size)\n", 246 | " weighted_average = True" 247 | ] 248 | }, 249 | { 250 | "cell_type": "code", 251 | "execution_count": null, 252 | "metadata": { 253 | "_uuid": "e4551ed97d6ee33f91283f559bbb7d9549d2d1f6" 254 | }, 255 | "outputs": [], 256 | "source": [ 257 | "train_gen,valid_gen = create_generators(args,b.preprocess_image)" 258 | ] 259 | }, 260 | { 261 | "cell_type": "markdown", 262 | "metadata": { 263 | "_uuid": "48482aa28034ede1fc5dd69661bbf83163ddb7bd" 264 | }, 265 | "source": [ 266 | "# Image Augmentation\n", 267 | "\n", 268 | "In addition to augmentations already done by keras-retinanet [here](https://github.com/fizyr/keras-retinanet/blob/master/keras_retinanet/bin/train.py#L227) , we'll use a package called imgaug to furthur augment the data.\n" 269 | ] 270 | }, 271 | { 272 | "cell_type": "code", 273 | "execution_count": null, 274 | "metadata": { 275 | "_uuid": "c09f817a02a3c385181a216f1c7c1af993efcf09" 276 | }, 277 | "outputs": [], 278 | "source": [ 279 | "sometimes = lambda aug: iaa.Sometimes(0.5, aug)\n", 280 | "# Define our sequence of augmentation steps that will be applied to every image.\n", 281 | "seq = iaa.Sequential(\n", 282 | " [\n", 283 | " #\n", 284 | " # Execute 1 to 9 of the following (less important) augmenters per\n", 285 | " # image. Don't execute all of them, as that would often be way too\n", 286 | " # strong.\n", 287 | " #\n", 288 | " iaa.SomeOf((1, 9),\n", 289 | " [\n", 290 | "\n", 291 | " # Blur each image with varying strength using\n", 292 | " # gaussian blur (sigma between 0 and .5),\n", 293 | " # average/uniform blur (kernel size 1x1)\n", 294 | " # median blur (kernel size 1x1).\n", 295 | " iaa.OneOf([\n", 296 | " iaa.GaussianBlur((0,0.5)),\n", 297 | " iaa.AverageBlur(k=(1)),\n", 298 | " iaa.MedianBlur(k=(1)),\n", 299 | " ]),\n", 300 | "\n", 301 | " # Sharpen each image, overlay the result with the original\n", 302 | " # image using an alpha between 0 (no sharpening) and 1\n", 303 | " # (full sharpening effect).\n", 304 | " iaa.Sharpen(alpha=(0, 0.25), lightness=(0.75, 1.5)),\n", 305 | "\n", 306 | " # Add gaussian noise to some images.\n", 307 | " # In 50% of these cases, the noise is randomly sampled per\n", 308 | " # channel and pixel.\n", 309 | " # In the other 50% of all cases it is sampled once per\n", 310 | " # pixel (i.e. brightness change).\n", 311 | " iaa.AdditiveGaussianNoise(\n", 312 | " loc=0, scale=(0.0, 0.01*255), per_channel=0.5\n", 313 | " ),\n", 314 | "\n", 315 | " # Either drop randomly 1 to 10% of all pixels (i.e. set\n", 316 | " # them to black) or drop them on an image with 2-5% percent\n", 317 | " # of the original size, leading to large dropped\n", 318 | " # rectangles.\n", 319 | " iaa.OneOf([\n", 320 | " iaa.Dropout((0.01, 0.1), per_channel=0.5),\n", 321 | " iaa.CoarseDropout(\n", 322 | " (0.03, 0.15), size_percent=(0.02, 0.05),\n", 323 | " per_channel=0.2\n", 324 | " ),\n", 325 | " ]),\n", 326 | "\n", 327 | " # Add a value of -5 to 5 to each pixel.\n", 328 | " iaa.Add((-5, 5), per_channel=0.5),\n", 329 | "\n", 330 | " # Change brightness of images (85-115% of original value).\n", 331 | " iaa.Multiply((0.85, 1.15), per_channel=0.5),\n", 332 | "\n", 333 | " # Improve or worsen the contrast of images.\n", 334 | " iaa.ContrastNormalization((0.75, 1.25), per_channel=0.5),\n", 335 | "\n", 336 | " # Convert each image to grayscale and then overlay the\n", 337 | " # result with the original with random alpha. I.e. remove\n", 338 | " # colors with varying strengths.\n", 339 | " iaa.Grayscale(alpha=(0.0, 0.25)),\n", 340 | "\n", 341 | " # In some images distort local areas with varying strength.\n", 342 | " sometimes(iaa.PiecewiseAffine(scale=(0.001, 0.01)))\n", 343 | " ],\n", 344 | " # do all of the above augmentations in random order\n", 345 | " random_order=True\n", 346 | " )\n", 347 | " ],\n", 348 | " # do all of the above augmentations in random order\n", 349 | " random_order=True\n", 350 | ")" 351 | ] 352 | }, 353 | { 354 | "cell_type": "code", 355 | "execution_count": null, 356 | "metadata": { 357 | "_uuid": "e2743e4a77f3b2d2e1c702b0c968e07e7603d2cc" 358 | }, 359 | "outputs": [], 360 | "source": [ 361 | "def augment_train_gen(train_gen,visualize=False):\n", 362 | " '''\n", 363 | " Creates a generator using another generator with applied image augmentation.\n", 364 | " Args\n", 365 | " train_gen : keras-retinanet generator object.\n", 366 | " visualize : Boolean; False will convert bounding boxes to their anchor box targets for the model.\n", 367 | " '''\n", 368 | " imgs = []\n", 369 | " boxes = []\n", 370 | " targets = []\n", 371 | " size = train_gen.size()\n", 372 | " idx = 0\n", 373 | " while True:\n", 374 | " while len(imgs) < args.batch_size:\n", 375 | " image = train_gen.load_image(idx % size)\n", 376 | " annotations = train_gen.load_annotations(idx % size)\n", 377 | " image,annotations = train_gen.random_transform_group_entry(image,annotations)\n", 378 | " imgs.append(image) \n", 379 | " boxes.append(annotations['bboxes'])\n", 380 | " targets.append(annotations)\n", 381 | " idx += 1\n", 382 | " if visualize:\n", 383 | " imgs = seq.augment_images(imgs)\n", 384 | " imgs = np.array(imgs)\n", 385 | " boxes = np.array(boxes)\n", 386 | " yield imgs,boxes\n", 387 | " else:\n", 388 | " imgs = seq.augment_images(imgs)\n", 389 | " imgs,targets = train_gen.preprocess_group(imgs,targets)\n", 390 | " imgs = train_gen.compute_inputs(imgs)\n", 391 | " targets = train_gen.compute_targets(imgs,targets)\n", 392 | " imgs = np.array(imgs)\n", 393 | " yield imgs,targets\n", 394 | " imgs = []\n", 395 | " boxes = []\n", 396 | " targets = []\n", 397 | " " 398 | ] 399 | }, 400 | { 401 | "cell_type": "markdown", 402 | "metadata": { 403 | "_uuid": "755f6bb05e3251d40877ddba74c8f6486678f47c" 404 | }, 405 | "source": [ 406 | "# Visualize augmentations" 407 | ] 408 | }, 409 | { 410 | "cell_type": "code", 411 | "execution_count": null, 412 | "metadata": { 413 | "_uuid": "bcd7d99112b845fe8b42ca980d3842361db93614" 414 | }, 415 | "outputs": [], 416 | "source": [ 417 | "import matplotlib.pyplot as plt\n", 418 | "\n", 419 | "skip_batches = 5\n", 420 | "i = 0\n", 421 | "\n", 422 | "for imgs,boxes in augment_train_gen(train_gen,visualize=True):\n", 423 | " if i > skip_batches:\n", 424 | " fig=plt.figure(figsize=(24,96))\n", 425 | " columns = 2\n", 426 | " rows = 8\n", 427 | " for i in range(1, columns*rows + 1):\n", 428 | " draw_boxes(imgs[i], boxes[i], (0, 255, 0), thickness=1)\n", 429 | " fig.add_subplot(rows, columns, i)\n", 430 | " plt.imshow(cv2.cvtColor(imgs[i],cv2.COLOR_BGR2RGB))\n", 431 | " plt.show()\n", 432 | " break\n", 433 | " else:\n", 434 | " i += 1\n" 435 | ] 436 | }, 437 | { 438 | "cell_type": "markdown", 439 | "metadata": { 440 | "_uuid": "35f45954269581be05dbc6a9601965db625609d8" 441 | }, 442 | "source": [ 443 | "# More Hyperparameters\n", 444 | "\n", 445 | "we'll use learning rate of 0.001 and freeze weights for the backbone" 446 | ] 447 | }, 448 | { 449 | "cell_type": "code", 450 | "execution_count": null, 451 | "metadata": { 452 | "_uuid": "de392d655a630d3b73dfa84fe1febbed0f3fdbe2" 453 | }, 454 | "outputs": [], 455 | "source": [ 456 | "model, training_model, prediction_model = create_models(\n", 457 | " backbone_retinanet=b.retinanet,\n", 458 | " num_classes=train_gen.num_classes(),\n", 459 | " weights=None,\n", 460 | " multi_gpu=False,\n", 461 | " freeze_backbone=True,\n", 462 | " lr=1e-3,\n", 463 | " config=args.config\n", 464 | " )" 465 | ] 466 | }, 467 | { 468 | "cell_type": "code", 469 | "execution_count": null, 470 | "metadata": { 471 | "_uuid": "a34e90ba8fec837af9007610241b487b9a6a807c" 472 | }, 473 | "outputs": [], 474 | "source": [ 475 | "callbacks = create_callbacks(\n", 476 | " model,\n", 477 | " training_model,\n", 478 | " prediction_model,\n", 479 | " valid_gen,\n", 480 | " args,\n", 481 | ")" 482 | ] 483 | }, 484 | { 485 | "cell_type": "markdown", 486 | "metadata": { 487 | "_uuid": "33285a3b74dc8580797f86678e080cbfecd7c668" 488 | }, 489 | "source": [ 490 | "# Download pretrained model\n", 491 | "\n", 492 | "We download a pretrained model on COCO dataset and load it's weights, we'll skip loading the weights for the few last layers" 493 | ] 494 | }, 495 | { 496 | "cell_type": "code", 497 | "execution_count": null, 498 | "metadata": { 499 | "_uuid": "764a0430cc7c5f5f341bb3ba3bb89b699248481e" 500 | }, 501 | "outputs": [], 502 | "source": [ 503 | "!wget https://github.com/fizyr/keras-retinanet/releases/download/0.5.0/resnet50_coco_best_v2.1.0.h5" 504 | ] 505 | }, 506 | { 507 | "cell_type": "code", 508 | "execution_count": null, 509 | "metadata": { 510 | "_uuid": "275ddf407b82092a0ec8a514500c0987f918d59b" 511 | }, 512 | "outputs": [], 513 | "source": [ 514 | "training_model.load_weights('resnet50_coco_best_v2.1.0.h5',skip_mismatch=True,by_name=True)" 515 | ] 516 | }, 517 | { 518 | "cell_type": "markdown", 519 | "metadata": { 520 | "_uuid": "f7cd58d21715eb0ec3a9de5b30be8cea85915726" 521 | }, 522 | "source": [ 523 | "# Train the model\n", 524 | "\n", 525 | "We will train for 70 epochs" 526 | ] 527 | }, 528 | { 529 | "cell_type": "code", 530 | "execution_count": null, 531 | "metadata": { 532 | "_uuid": "06bca3482f679c50640b7396bafb9b291b4f402b" 533 | }, 534 | "outputs": [], 535 | "source": [ 536 | "training_model.fit_generator(generator=augment_train_gen(train_gen),\n", 537 | " steps_per_epoch=args.steps,\n", 538 | " epochs=args.epochs,\n", 539 | " verbose=1,\n", 540 | " callbacks=callbacks,)" 541 | ] 542 | }, 543 | { 544 | "cell_type": "code", 545 | "execution_count": null, 546 | "metadata": { 547 | "_uuid": "7a5b84c37c58444c37e6d08fb6de19b756c03926" 548 | }, 549 | "outputs": [], 550 | "source": [] 551 | } 552 | ], 553 | "metadata": { 554 | "kernelspec": { 555 | "display_name": "Python 3", 556 | "language": "python", 557 | "name": "python3" 558 | }, 559 | "language_info": { 560 | "codemirror_mode": { 561 | "name": "ipython", 562 | "version": 3 563 | }, 564 | "file_extension": ".py", 565 | "mimetype": "text/x-python", 566 | "name": "python", 567 | "nbconvert_exporter": "python", 568 | "pygments_lexer": "ipython3", 569 | "version": "3.6.7" 570 | } 571 | }, 572 | "nbformat": 4, 573 | "nbformat_minor": 2 574 | } 575 | -------------------------------------------------------------------------------- /visualize.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "markdown", 5 | "metadata": {}, 6 | "source": [ 7 | "# Visualize predictions on test set\n", 8 | "\n", 9 | "Set image_id to visualize predictions for that image" 10 | ] 11 | }, 12 | { 13 | "cell_type": "code", 14 | "execution_count": null, 15 | "metadata": {}, 16 | "outputs": [], 17 | "source": [ 18 | "from keras_retinanet.utils.visualization import draw_box, draw_caption\n", 19 | "from keras_retinanet.utils.colors import label_color\n", 20 | "from keras_retinanet.utils.image import read_image_bgr, preprocess_image, resize_image\n", 21 | "import cv2\n", 22 | "from os import walk\n", 23 | "import matplotlib.pyplot as plt\n", 24 | "\n", 25 | "_,_,image_ids = next(walk('test_data_images/images'))\n", 26 | "image_ids = [i[:-4] for i in image_ids]\n", 27 | "image_ids = sorted(image_ids)\n", 28 | "\n", 29 | "idx = 0\n", 30 | "image_id = 1\n", 31 | "score_thres = 0.\n", 32 | "\n", 33 | "for id in image_ids:\n", 34 | " # load image\n", 35 | " idx += 1\n", 36 | " if idx == image_id:\n", 37 | " image = read_image_bgr('test_data_images/images/' + id + '.jpg')\n", 38 | "\n", 39 | " # copy to draw on\n", 40 | " draw = image.copy()\n", 41 | " draw = cv2.cvtColor(draw, cv2.COLOR_BGR2RGB)\n", 42 | "\n", 43 | " # process image\n", 44 | " boxes = [list(map(int,(line.split()[3],line.split()[2],line.split()[5],line.split()[4]))) for line in open('predictions/' + id + '.txt','r').readlines()]\n", 45 | " scores = [float(line.split()[1]) for line in open('predictions/' + id + '.txt','r').readlines()]\n", 46 | " labels = [int(line.split()[0]) - 1 for line in open('predictions/' + id + '.txt','r').readlines()]\n", 47 | " for box, score, label in zip(boxes, scores, labels):\n", 48 | " if score < score_thres:\n", 49 | " break\n", 50 | " color = label_color(label)\n", 51 | " draw_box(draw, box, color=color,thickness=1)\n", 52 | " caption = \"{:.3f}\".format(score)\n", 53 | " draw_caption(draw, box, caption)\n", 54 | "\n", 55 | " plt.figure(figsize=(15, 15))\n", 56 | " plt.axis('off')\n", 57 | " plt.imshow(draw)\n", 58 | " plt.show()\n", 59 | " break" 60 | ] 61 | }, 62 | { 63 | "cell_type": "code", 64 | "execution_count": null, 65 | "metadata": {}, 66 | "outputs": [], 67 | "source": [] 68 | } 69 | ], 70 | "metadata": { 71 | "kernelspec": { 72 | "display_name": "Python 3", 73 | "language": "python", 74 | "name": "python3" 75 | }, 76 | "language_info": { 77 | "codemirror_mode": { 78 | "name": "ipython", 79 | "version": 3 80 | }, 81 | "file_extension": ".py", 82 | "mimetype": "text/x-python", 83 | "name": "python", 84 | "nbconvert_exporter": "python", 85 | "pygments_lexer": "ipython3", 86 | "version": "3.6.7" 87 | } 88 | }, 89 | "nbformat": 4, 90 | "nbformat_minor": 2 91 | } 92 | --------------------------------------------------------------------------------