├── .gitignore ├── LICENSE ├── README.md ├── archive ├── znz-classify-buildings-20181119.ipynb ├── znz-classify-buildings-20181206.ipynb ├── znz-classify-buildings-20190118.ipynb ├── znz-eval-20190118.ipynb ├── znz-inference-20190118.ipynb ├── znz-postprocess-20190118.ipynb ├── znz-segment-buildingfootprint-20181119.ipynb ├── znz-segment-buildingfootprint-20181129-comboloss-rn34.ipynb ├── znz-segment-buildingfootprint-20181205-comboloss-rn34.ipynb └── znz-segment-buildingfootprint-20190108-alldata.ipynb ├── geo_fastai_tutorial01_public_v1.ipynb └── static ├── grid119_preview.png ├── overview_predict.png ├── overview_train.png └── znz-demo.gif /.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 | build/ 12 | develop-eggs/ 13 | dist/ 14 | downloads/ 15 | eggs/ 16 | .eggs/ 17 | lib/ 18 | lib64/ 19 | parts/ 20 | sdist/ 21 | var/ 22 | wheels/ 23 | *.egg-info/ 24 | .installed.cfg 25 | *.egg 26 | MANIFEST 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 | .pytest_cache/ 49 | 50 | # Translations 51 | *.mo 52 | *.pot 53 | 54 | # Django stuff: 55 | *.log 56 | local_settings.py 57 | db.sqlite3 58 | 59 | # Flask stuff: 60 | instance/ 61 | .webassets-cache 62 | 63 | # Scrapy stuff: 64 | .scrapy 65 | 66 | # Sphinx documentation 67 | docs/_build/ 68 | 69 | # PyBuilder 70 | target/ 71 | 72 | # Jupyter Notebook 73 | .ipynb_checkpoints 74 | 75 | # pyenv 76 | .python-version 77 | 78 | # celery beat schedule file 79 | celerybeat-schedule 80 | 81 | # SageMath parsed files 82 | *.sage.py 83 | 84 | # Environments 85 | .env 86 | .venv 87 | env/ 88 | venv/ 89 | ENV/ 90 | env.bak/ 91 | venv.bak/ 92 | 93 | # Spyder project settings 94 | .spyderproject 95 | .spyproject 96 | 97 | # Rope project settings 98 | .ropeproject 99 | 100 | # mkdocs documentation 101 | /site 102 | 103 | # mypy 104 | .mypy_cache/ 105 | 106 | .ipynb_checkpoints/ 107 | 108 | *.zip 109 | *.DS_Store 110 | readme_draft.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 | # Aerial Mapping with Drones & Deep Learning in Zanzibar, Tanzania 2 | 3 | ## Motivation: 4 | 5 | Open source R&D notebooks of all the steps (deep learning and otherwise) to create a state of the art deep learning building detector & classifier from high-resolution aerial/drone imagery. Something like this: 6 | 7 | ### Interactive version: [https://alpha.anthropo.co/znz-demo](https://alpha.anthropo.co/znz-demo) 8 | ![static/znz-demo.gif](static/znz-demo.gif) 9 | 10 | 11 | ## 7/25/2019 Update: 12 | 13 | In process of rewriting everything as a series of interactive geospatial deep learning tutorials on [Google Colab](https://colab.research.google.com/). 14 | 15 | **See 1st tutorial published 7/25/2019** for a complete data creation, model creation, inference, and evaluation workflow for building segmentation: 16 | 17 | - [Medium post](https://medium.com/@anthropoco/how-to-segment-buildings-on-drone-imagery-with-fast-ai-cloud-native-geodata-tools-ae249612c321?source=friends_link&sk=57b82002ac47724ecf9a2aaa98de994b) 18 | 19 | - [Previewable notebook](https://nbviewer.jupyter.org/github/daveluo/zanzibar-aerial-mapping/blob/master/geo_fastai_tutorial01_public_v1.ipynb) 20 | 21 | - [Open in Colab](https://colab.research.google.com/github/daveluo/zanzibar-aerial-mapping/blob/master/geo_fastai_tutorial01_public_v1.ipynb) 22 | 23 | Prior dev notebooks can be found in /archive with details preserved below the line: 24 | 25 | 26 | ------------------------ 27 | 28 | 29 | ## Results: 30 | 31 | As of 1/18/2019 (internal val only): 32 | 33 | | | Mean F1 score | Foundation F1 | Unfinished F1 | Completed F1 | All Buildings F1 | 34 | |-----------------------------|---------------|---------------|---------------|--------------|------------------| 35 | | Internal Val (grid 042) | 0.728 | 0.718 | 0.755 | 0.710 | 0.796 | 36 | 37 | 38 | [Top 2 in WeRobotics' Open AI Tanzania Challenge](https://blog.werobotics.org/2018/12/06/announcing-the-winners-of-the-open-ai-tanzania-challenge/) 39 | 40 | 41 | | | Mean F1 score | Foundation F1 | Unfinished F1 | Completed F1 | All Buildings F1 | 42 | |-----------------------------|---------------|---------------|---------------|--------------|------------------| 43 | | Final Test (grids 059, 066) | 0.697 | 0.744 | 0.692 | 0.655 | 0.723 | 44 | | Internal Val (grid 042) | 0.696 | 0.683 | 0.749 | 0.656 | 0.757 | 45 | 46 | 47 | 48 | ## Background: 49 | 50 | https://blog.werobotics.org/2018/08/06/welcome-to-the-open-ai-tanzania-challenge/ 51 | 52 | > Maps are absolutely essential for decision support. Knowing where buildings are located is a fundamental input for urban planning, public safety, public health, disaster response, environmental protection, sustainable development and census data, for example. Some of these applications typically require timely and high-resolution maps. 53 | 54 | ![https://blog.werobotics.org/wp-content/uploads/2018/08/zmi-geonode.png](https://blog.werobotics.org/wp-content/uploads/2018/08/zmi-geonode.png) 55 | 56 | > Take the following scenario: a local organization that provides low-cost solar panels to low-income households in rural Tanzania is evaluating a large neighborhood with many small houses. They have to determine how to best optimize the installation and distribution of their panels. So they need to know which of the residential structures are oriented in a way that makes them more suitable for solar panels. Knowing where these small houses are and what their orientation is vis-a-vis the sun and surrounding trees, what their roofs look like to determine where to place the panels, and what materials the roofs are made of are all key inputs for their planning. This is just one of many applications for high-resolution maps. 57 | 58 | 59 | 60 | https://competitions.codalab.org/competitions/20100#learn_the_details 61 | > Open AI Tanzania — is a partnership with our friends at the State University of Zanzibar (SUZA), WeRobotics, World Bank, OpenAerialMap and Tanzania Flying Labs. Open AI Tanzania invites data scientists to develop feature detection algorithms that can automatically identify buildings and building types using high-resolution aerial imagery collected by Tanzanian drone pilots through the Zanzibar Mapping Initiative (ZMI). The goal of this challenge is to correctly segment and classify building footprints under various stages of construction. 62 | 63 | 64 | ## Source imagery & training data license 65 | 66 | > We request that all participants fill out a [Google Form](https://docs.google.com/forms/d/e/1FAIpQLSewpoY650nUHyl5kobIWl68Msk2QFBEC8XFCAV6lZSwbVdqUw/viewform). 67 | 68 | > The imagery data is released as OpenData using the Creative Commons Attribution 4.0 International license, attribution must be given to: Commission for Lands (COLA), Revolutionary Government of Zanzibar (RGoZ) 69 | 70 | ## Overview: 71 | 72 | ### Training workflow: 73 | 74 | ![static/overview_train.png](static/overview_train.png) 75 | 76 | ### Prediction workflow: 77 | 78 | ![static/overview_predict.png](static/overview_predict.png) 79 | 80 | ## Notebooks (7/25/19 moved to archive): 81 | 82 | ### [archive/znz-segment-buildingfootprint-20190108-alldata.ipynb](archive/znz-segment-buildingfootprint-20190108-alldata.ipynb) 83 | 84 | - segmentation model for pixel-level mapping of every building structure, regardless of condition 85 | - trained on image chips at three zooms: z20, z19, z18 86 | - combined BCE/dice loss function, pretrained resnet34 encoder 87 | - dice: 0.863, accuracy: 98.1% 88 | 89 | ### [archive/znz-classify-buildings-20190118.ipynb](archive/znz-classify-buildings-20190118.ipynb) 90 | 91 | - building classification by condition (Complete, Incomplete, Foundation, Empty) after detection/segmentation 92 | - BCE loss, pretrained resnet50 93 | - accuracy: 94% 94 | 95 | ### [archive/znz-inference-20190118.ipynb](archive/znz-inference-20190118.ipynb) 96 | 97 | - windowed reads and sub-windowed reads with rasterio to run inference on cloud-optimized geotiffs (COG) of arbitirary sizes 98 | - merge back to full cog extent 99 | 100 | ### [archive/znz-postprocess-20190118.ipynb](archive/znz-postprocess-20190118.ipynb) 101 | 102 | - thresholding, polygonization of windowed reads, deduping, save as geojson 103 | - creating detected building crops for classifier 104 | - updating geojson with class predictions 105 | 106 | ### [archive/znz-eval-20190118.ipynb](archive/znz-eval-20190118.ipynb) 107 | 108 | - evaluation scripts for precision, recall, F1 score 109 | - adapted from spacenet utilities: https://github.com/SpaceNetChallenge/utilities/tree/master/python 110 | 111 | ## Ready-to-train preprocessed datasets 112 | 113 | ### [znz-segment-z19.zip](https://www.dropbox.com/s/v1zvgrv06alogkk/znz-segment-z19.zip?dl=0) (212 MB): 114 | 115 | - 2,691 512x512 square image chips at zoom level 19 (~0.3m/pixel) as .jpg files with corresponding binary masks as .png files 116 | - Unzips as znz-train-z19-all-buffered/images-512 and znz-train-z19-all-buffered/masks-512 117 | 118 | ### [znz-classify.zip](https://www.dropbox.com/s/9ge0a2kpuv0a0lk/znz-classify.zip?dl=0) (259 MB): 119 | 120 | - 20,176 images of various sizes & shapes labeled as 4 classes: Complete, Incomplete, Foundation, Empty 121 | - Unzips as /images 122 | 123 | 124 | ## Thanks 125 | 126 | - Commission for Lands (COLA), Revolutionary Government of Zanzibar (RGoZ) 127 | - Zanzibar Mapping Initiative 128 | - WeRobotics 129 | - OpenAerialMap 130 | - Fast.ai team and library v1 131 | - Fast.ai forum participants and fellow students (particularly @KarlH, @henripal for code/nbs) 132 | 133 | ## TO DO 134 | 135 | - [x] train segmentation model in fastai v1 up to or exceeding prior performance level of dice = 0.847, accuracy = 0.977 from fastai v0.7 136 | - [x] polygonization nb 137 | - [x] prediction thresholding and clean-up nb 138 | - [x] evaluation scripts 139 | - [x] training image mask/tile creation nb 140 | -------------------------------------------------------------------------------- /archive/znz-eval-20190118.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "markdown", 5 | "metadata": {}, 6 | "source": [ 7 | "Eval scripts adapted from https://github.com/SpaceNetChallenge/utilities/tree/master/python" 8 | ] 9 | }, 10 | { 11 | "cell_type": "code", 12 | "execution_count": 1, 13 | "metadata": { 14 | "collapsed": true 15 | }, 16 | "outputs": [], 17 | "source": [ 18 | "import numpy as np\n", 19 | "import geopandas as gpd\n", 20 | "import rtree\n", 21 | "\n", 22 | "from pathlib import Path\n", 23 | "\n", 24 | "import matplotlib.pyplot as plt\n", 25 | "import matplotlib as mpl\n", 26 | "%matplotlib inline\n", 27 | "\n", 28 | "from tqdm import tqdm" 29 | ] 30 | }, 31 | { 32 | "cell_type": "code", 33 | "execution_count": 2, 34 | "metadata": { 35 | "collapsed": true 36 | }, 37 | "outputs": [], 38 | "source": [ 39 | "def create_rtree_from_poly(poly_list):\n", 40 | " # create index\n", 41 | " index = rtree.index.Index(interleaved=False)\n", 42 | " for idx, building in enumerate(poly_list):\n", 43 | " minx, miny, maxx, maxy = building.bounds\n", 44 | " envelope = (minx, maxx, miny, maxy)\n", 45 | " index.insert(idx, envelope)\n", 46 | "\n", 47 | " return index\n", 48 | "\n", 49 | "def search_rtree(test_building, index):\n", 50 | " # input test poly ogr.Geometry and rtree index\n", 51 | " if test_building.type == 'Polygon' or \\\n", 52 | " test_building.type == 'MultiPolygon':\n", 53 | " minx, miny, maxx, maxy = test_building.bounds\n", 54 | " envelope = (minx, maxx, miny, maxy) \n", 55 | " fidlist = index.intersection(envelope)\n", 56 | " else:\n", 57 | " fidlist = []\n", 58 | "\n", 59 | " return fidlist\n" 60 | ] 61 | }, 62 | { 63 | "cell_type": "code", 64 | "execution_count": 3, 65 | "metadata": { 66 | "collapsed": true 67 | }, 68 | "outputs": [], 69 | "source": [ 70 | "def iou(test_poly, truth_polys, truth_index=[]):\n", 71 | " fidlistArray = []\n", 72 | " iou_list = []\n", 73 | " \n", 74 | " if truth_index:\n", 75 | " fidlist = search_rtree(test_poly, truth_index)\n", 76 | "\n", 77 | " for fid in fidlist:\n", 78 | " if not test_poly.is_valid:\n", 79 | " test_poly = test_poly.buffer(0.0)\n", 80 | "\n", 81 | " intersection_result = test_poly.intersection(truth_polys[fid].buffer(0.0))\n", 82 | " fidlistArray.append(fid)\n", 83 | "\n", 84 | " if intersection_result.type == 'Polygon' or \\\n", 85 | " intersection_result.type == 'MultiPolygon':\n", 86 | " intersection_area = intersection_result.area\n", 87 | " union_area = test_poly.union(truth_polys[fid].buffer(0.0)).area\n", 88 | " iou_list.append(intersection_area / union_area)\n", 89 | "\n", 90 | " else:\n", 91 | " iou_list.append(0)\n", 92 | "\n", 93 | " else:\n", 94 | " for idx, truth_poly in enumerate(truth_polys):\n", 95 | " if not test_poly.is_valid or not truth_poly.is_valid:\n", 96 | " test_poly = test_poly.buffer(0.0)\n", 97 | " truth_poly = truth_poly.buffer(0.0)\n", 98 | "# print(f'fixed geom error at {idx}')\n", 99 | "\n", 100 | " intersection_result = test_poly.intersection(truth_poly)\n", 101 | " #print(idx, intersection_result.type)\n", 102 | "\n", 103 | " if intersection_result.type == 'Polygon' or \\\n", 104 | " intersection_result.type == 'MultiPolygon':\n", 105 | " intersection_area = intersection_result.area\n", 106 | " union_area = test_poly.union(truth_poly).area\n", 107 | " iou_list.append(intersection_area / union_area)\n", 108 | " # print(f'found intersect at test_poly {i} with truth poly {idx}')\n", 109 | " # print(intersection_area/union_area)\n", 110 | " else: \n", 111 | " iou_list.append(0)\n", 112 | " \n", 113 | " return iou_list, fidlistArray" 114 | ] 115 | }, 116 | { 117 | "cell_type": "code", 118 | "execution_count": 4, 119 | "metadata": { 120 | "collapsed": true 121 | }, 122 | "outputs": [], 123 | "source": [ 124 | "def score(test_polys, truth_polys, threshold=0.5, truth_index=[],\n", 125 | " resultGeoJsonName = [],\n", 126 | " imageId = []):\n", 127 | "\n", 128 | " # Define internal functions\n", 129 | "\n", 130 | " # Find detections using threshold/argmax/IoU for test polygons\n", 131 | " true_pos_count = 0\n", 132 | " false_pos_count = 0\n", 133 | " truth_poly_count = len(truth_polys)\n", 134 | " \n", 135 | " true_ids = []\n", 136 | " false_ids = []\n", 137 | "\n", 138 | " for idx, test_poly in tqdm(enumerate(test_polys)):\n", 139 | " if truth_polys:\n", 140 | " iou_list, fidlist = iou(test_poly, truth_polys, truth_index)\n", 141 | " if not iou_list:\n", 142 | " maxiou = 0\n", 143 | " else:\n", 144 | " maxiou = np.max(iou_list)\n", 145 | "\n", 146 | "# print(maxiou, iou_list, fidlist)\n", 147 | " if maxiou >= threshold:\n", 148 | " true_pos_count += 1\n", 149 | " true_ids.append(idx)\n", 150 | " minx, miny, maxx, maxy = truth_polys[fidlist[np.argmax(iou_list)]].bounds\n", 151 | " envelope = (minx, maxx, miny, maxy) \n", 152 | " truth_index.delete(fidlist[np.argmax(iou_list)], envelope)\n", 153 | " #del truth_polys[fidlist[np.argmax(iou_list)]]\n", 154 | " else:\n", 155 | " false_pos_count += 1\n", 156 | " false_ids.append(idx)\n", 157 | " else:\n", 158 | " false_pos_count += 1\n", 159 | " false_ids.append(idx)\n", 160 | "\n", 161 | " false_neg_count = truth_poly_count - true_pos_count\n", 162 | "\n", 163 | " return true_pos_count, false_pos_count, false_neg_count, true_ids, false_ids" 164 | ] 165 | }, 166 | { 167 | "cell_type": "code", 168 | "execution_count": 5, 169 | "metadata": { 170 | "collapsed": true 171 | }, 172 | "outputs": [], 173 | "source": [ 174 | "def evalfunction(image_id, test_polys, truth_polys, truth_index=[], resultGeoJsonName=[], threshold = 0.5):\n", 175 | "\n", 176 | " if len(truth_polys)==0:\n", 177 | " true_pos_count = 0\n", 178 | " false_pos_count = len(test_polys)\n", 179 | " false_neg_count = 0\n", 180 | " else:\n", 181 | " true_pos_count, false_pos_count, false_neg_count, true_ids, false_ids = score(test_polys, truth_polys,\n", 182 | " truth_index=truth_index,\n", 183 | " resultGeoJsonName=resultGeoJsonName,\n", 184 | " imageId=image_id,\n", 185 | " threshold=threshold\n", 186 | " )\n", 187 | "\n", 188 | "\n", 189 | " if (true_pos_count > 0):\n", 190 | "\n", 191 | " precision = float(true_pos_count) / (float(true_pos_count) + float(false_pos_count))\n", 192 | " recall = float(true_pos_count) / (float(true_pos_count) + float(false_neg_count))\n", 193 | " F1score = 2.0 * precision * recall / (precision + recall)\n", 194 | " else:\n", 195 | " F1score = 0\n", 196 | " return ((F1score, true_pos_count, false_pos_count, false_neg_count), true_ids, false_ids, image_id)" 197 | ] 198 | }, 199 | { 200 | "cell_type": "code", 201 | "execution_count": 6, 202 | "metadata": { 203 | "collapsed": true 204 | }, 205 | "outputs": [], 206 | "source": [ 207 | "def precision_recall(true_pos_count, false_pos_count, false_neg_count):\n", 208 | " precision = float(true_pos_count) / (float(true_pos_count) + float(false_pos_count))\n", 209 | " recall = float(true_pos_count) / (float(true_pos_count) + float(false_neg_count))\n", 210 | " return (precision, recall)" 211 | ] 212 | }, 213 | { 214 | "cell_type": "code", 215 | "execution_count": 7, 216 | "metadata": { 217 | "collapsed": true 218 | }, 219 | "outputs": [], 220 | "source": [ 221 | "TRUTH = Path('znz-input')\n", 222 | "TEST = Path('znz-20190118')" 223 | ] 224 | }, 225 | { 226 | "cell_type": "code", 227 | "execution_count": 95, 228 | "metadata": { 229 | "collapsed": true 230 | }, 231 | "outputs": [], 232 | "source": [ 233 | "df_truth = gpd.read_file(f'{str(TRUTH)}/grid_042.geojson')\n", 234 | "df_test = gpd.read_file(f'{str(TEST)}/grid_042_20190118_07_classes.geojson')" 235 | ] 236 | }, 237 | { 238 | "cell_type": "code", 239 | "execution_count": 96, 240 | "metadata": {}, 241 | "outputs": [ 242 | { 243 | "data": { 244 | "text/html": [ 245 | "
\n", 246 | "\n", 259 | "\n", 260 | " \n", 261 | " \n", 262 | " \n", 263 | " \n", 264 | " \n", 265 | " \n", 266 | " \n", 267 | " \n", 268 | " \n", 269 | " \n", 270 | " \n", 271 | " \n", 272 | " \n", 273 | " \n", 274 | " \n", 275 | " \n", 276 | " \n", 277 | " \n", 278 | " \n", 279 | " \n", 280 | " \n", 281 | " \n", 282 | " \n", 283 | " \n", 284 | " \n", 285 | " \n", 286 | " \n", 287 | " \n", 288 | " \n", 289 | " \n", 290 | " \n", 291 | " \n", 292 | " \n", 293 | " \n", 294 | " \n", 295 | " \n", 296 | " \n", 297 | " \n", 298 | " \n", 299 | " \n", 300 | " \n", 301 | " \n", 302 | " \n", 303 | " \n", 304 | " \n", 305 | " \n", 306 | " \n", 307 | " \n", 308 | " \n", 309 | " \n", 310 | " \n", 311 | " \n", 312 | " \n", 313 | " \n", 314 | " \n", 315 | " \n", 316 | " \n", 317 | " \n", 318 | "
idchangesetproblematiconditionareageometry
012017-09-03T23:40:59NoneComplete16.93017578125POLYGON ((39.33633038681167 -5.920836943343485...
122017-09-03T23:40:59NoneComplete21.62890625POLYGON ((39.33628382960306 -5.92092792350677,...
232017-09-03T23:40:59NoneComplete11.2216796875POLYGON ((39.33622587109281 -5.920941122736775...
342017-09-03T23:40:59NoneComplete46.849609375POLYGON ((39.33368444829663 -5.923518950659385...
452017-09-03T23:40:59NoneComplete7.458984375POLYGON ((39.33414129839756 -5.923375141989136...
\n", 319 | "
" 320 | ], 321 | "text/plain": [ 322 | " id changeset problemati condition area \\\n", 323 | "0 1 2017-09-03T23:40:59 None Complete 16.93017578125 \n", 324 | "1 2 2017-09-03T23:40:59 None Complete 21.62890625 \n", 325 | "2 3 2017-09-03T23:40:59 None Complete 11.2216796875 \n", 326 | "3 4 2017-09-03T23:40:59 None Complete 46.849609375 \n", 327 | "4 5 2017-09-03T23:40:59 None Complete 7.458984375 \n", 328 | "\n", 329 | " geometry \n", 330 | "0 POLYGON ((39.33633038681167 -5.920836943343485... \n", 331 | "1 POLYGON ((39.33628382960306 -5.92092792350677,... \n", 332 | "2 POLYGON ((39.33622587109281 -5.920941122736775... \n", 333 | "3 POLYGON ((39.33368444829663 -5.923518950659385... \n", 334 | "4 POLYGON ((39.33414129839756 -5.923375141989136... " 335 | ] 336 | }, 337 | "execution_count": 96, 338 | "metadata": {}, 339 | "output_type": "execute_result" 340 | } 341 | ], 342 | "source": [ 343 | "df_truth.head()" 344 | ] 345 | }, 346 | { 347 | "cell_type": "code", 348 | "execution_count": 97, 349 | "metadata": {}, 350 | "outputs": [ 351 | { 352 | "data": { 353 | "text/html": [ 354 | "
\n", 355 | "\n", 368 | "\n", 369 | " \n", 370 | " \n", 371 | " \n", 372 | " \n", 373 | " \n", 374 | " \n", 375 | " \n", 376 | " \n", 377 | " \n", 378 | " \n", 379 | " \n", 380 | " \n", 381 | " \n", 382 | " \n", 383 | " \n", 384 | " \n", 385 | " \n", 386 | " \n", 387 | " \n", 388 | " \n", 389 | " \n", 390 | " \n", 391 | " \n", 392 | " \n", 393 | " \n", 394 | " \n", 395 | " \n", 396 | " \n", 397 | " \n", 398 | " \n", 399 | " \n", 400 | " \n", 401 | " \n", 402 | " \n", 403 | " \n", 404 | " \n", 405 | " \n", 406 | " \n", 407 | " \n", 408 | " \n", 409 | " \n", 410 | " \n", 411 | " \n", 412 | " \n", 413 | " \n", 414 | " \n", 415 | " \n", 416 | " \n", 417 | " \n", 418 | " \n", 419 | " \n", 420 | " \n", 421 | " \n", 422 | " \n", 423 | " \n", 424 | " \n", 425 | " \n", 426 | " \n", 427 | "
catbuilding_idconf_foundationconf_completedconf_unfinishedgeometry
0conf_completed00.00490.97650.0185POLYGON ((39.33541308434184 -5.933007141989886...
1conf_completed10.00960.62280.3669POLYGON ((39.33543994125422 -5.932787098088375...
2conf_completed20.00620.95250.0408POLYGON ((39.33536607110349 -5.932866080282638...
3conf_completed30.01890.97300.0079POLYGON ((39.33511558848227 -5.932883043772769...
4conf_foundation40.86730.02800.1034POLYGON ((39.33529523898535 -5.932884091309337...
\n", 428 | "
" 429 | ], 430 | "text/plain": [ 431 | " cat building_id conf_foundation conf_completed \\\n", 432 | "0 conf_completed 0 0.0049 0.9765 \n", 433 | "1 conf_completed 1 0.0096 0.6228 \n", 434 | "2 conf_completed 2 0.0062 0.9525 \n", 435 | "3 conf_completed 3 0.0189 0.9730 \n", 436 | "4 conf_foundation 4 0.8673 0.0280 \n", 437 | "\n", 438 | " conf_unfinished geometry \n", 439 | "0 0.0185 POLYGON ((39.33541308434184 -5.933007141989886... \n", 440 | "1 0.3669 POLYGON ((39.33543994125422 -5.932787098088375... \n", 441 | "2 0.0408 POLYGON ((39.33536607110349 -5.932866080282638... \n", 442 | "3 0.0079 POLYGON ((39.33511558848227 -5.932883043772769... \n", 443 | "4 0.1034 POLYGON ((39.33529523898535 -5.932884091309337... " 444 | ] 445 | }, 446 | "execution_count": 97, 447 | "metadata": {}, 448 | "output_type": "execute_result" 449 | } 450 | ], 451 | "source": [ 452 | "df_test.head()" 453 | ] 454 | }, 455 | { 456 | "cell_type": "code", 457 | "execution_count": 98, 458 | "metadata": {}, 459 | "outputs": [ 460 | { 461 | "data": { 462 | "text/plain": [ 463 | "" 464 | ] 465 | }, 466 | "execution_count": 98, 467 | "metadata": {}, 468 | "output_type": "execute_result" 469 | }, 470 | { 471 | "data": { 472 | "image/png": "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\n", 473 | "text/plain": [ 474 | "
" 475 | ] 476 | }, 477 | "metadata": {}, 478 | "output_type": "display_data" 479 | } 480 | ], 481 | "source": [ 482 | "df_truth.geometry.plot(figsize=(10,10))" 483 | ] 484 | }, 485 | { 486 | "cell_type": "code", 487 | "execution_count": 99, 488 | "metadata": {}, 489 | "outputs": [ 490 | { 491 | "data": { 492 | "text/plain": [ 493 | "" 494 | ] 495 | }, 496 | "execution_count": 99, 497 | "metadata": {}, 498 | "output_type": "execute_result" 499 | }, 500 | { 501 | "data": { 502 | "image/png": "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\n", 503 | "text/plain": [ 504 | "
" 505 | ] 506 | }, 507 | "metadata": {}, 508 | "output_type": "display_data" 509 | } 510 | ], 511 | "source": [ 512 | "df_test.geometry.plot(figsize=(10,10))" 513 | ] 514 | }, 515 | { 516 | "cell_type": "code", 517 | "execution_count": 100, 518 | "metadata": {}, 519 | "outputs": [ 520 | { 521 | "data": { 522 | "text/plain": [ 523 | "conf_completed 365\n", 524 | "conf_unfinished 108\n", 525 | "conf_foundation 51\n", 526 | "Name: cat, dtype: int64" 527 | ] 528 | }, 529 | "execution_count": 100, 530 | "metadata": {}, 531 | "output_type": "execute_result" 532 | } 533 | ], 534 | "source": [ 535 | "df_test['cat'].value_counts()" 536 | ] 537 | }, 538 | { 539 | "cell_type": "code", 540 | "execution_count": 101, 541 | "metadata": {}, 542 | "outputs": [ 543 | { 544 | "data": { 545 | "text/plain": [ 546 | "Complete 373\n", 547 | "Incomplete 112\n", 548 | "Foundation 66\n", 549 | "Name: condition, dtype: int64" 550 | ] 551 | }, 552 | "execution_count": 101, 553 | "metadata": {}, 554 | "output_type": "execute_result" 555 | } 556 | ], 557 | "source": [ 558 | "df_truth['condition'].value_counts()" 559 | ] 560 | }, 561 | { 562 | "cell_type": "code", 563 | "execution_count": 102, 564 | "metadata": { 565 | "collapsed": true 566 | }, 567 | "outputs": [], 568 | "source": [ 569 | "cats = [('conf_foundation','Foundation'),('conf_unfinished','Incomplete'),('conf_completed','Complete')]" 570 | ] 571 | }, 572 | { 573 | "cell_type": "code", 574 | "execution_count": 103, 575 | "metadata": {}, 576 | "outputs": [ 577 | { 578 | "name": "stderr", 579 | "output_type": "stream", 580 | "text": [ 581 | "51it [00:00, 372.86it/s]\n", 582 | "35it [00:00, 346.39it/s]" 583 | ] 584 | }, 585 | { 586 | "name": "stdout", 587 | "output_type": "stream", 588 | "text": [ 589 | "Foundation\n", 590 | "(0.717948717948718, 42, 9, 24) (0.8235294117647058, 0.6363636363636364)\n" 591 | ] 592 | }, 593 | { 594 | "name": "stderr", 595 | "output_type": "stream", 596 | "text": [ 597 | "108it [00:00, 394.50it/s]\n", 598 | "50it [00:00, 493.64it/s]" 599 | ] 600 | }, 601 | { 602 | "name": "stdout", 603 | "output_type": "stream", 604 | "text": [ 605 | "Incomplete\n", 606 | "(0.7545454545454546, 83, 25, 29) (0.7685185185185185, 0.7410714285714286)\n" 607 | ] 608 | }, 609 | { 610 | "name": "stderr", 611 | "output_type": "stream", 612 | "text": [ 613 | "365it [00:00, 442.09it/s]" 614 | ] 615 | }, 616 | { 617 | "name": "stdout", 618 | "output_type": "stream", 619 | "text": [ 620 | "Complete\n", 621 | "(0.7100271002710028, 262, 103, 111) (0.7178082191780822, 0.7024128686327078)\n" 622 | ] 623 | }, 624 | { 625 | "name": "stderr", 626 | "output_type": "stream", 627 | "text": [ 628 | "\n" 629 | ] 630 | } 631 | ], 632 | "source": [ 633 | "for (test_cat, truth_cat) in cats:\n", 634 | " test_polys = [geom for geom in df_test[df_test['cat'] == test_cat].geometry]\n", 635 | " truth_polys = [geom for geom in df_truth[df_truth['condition'] == truth_cat].geometry]\n", 636 | " truth_index = create_rtree_from_poly(truth_polys)\n", 637 | " scores = evalfunction(grid_num,test_polys, truth_polys, truth_index=truth_index)\n", 638 | " print(truth_cat)\n", 639 | " print(scores[0],precision_recall(*scores[0][1:]))" 640 | ] 641 | }, 642 | { 643 | "cell_type": "code", 644 | "execution_count": 104, 645 | "metadata": { 646 | "collapsed": true 647 | }, 648 | "outputs": [], 649 | "source": [ 650 | "test_polys = [geom for geom in df_test.geometry]\n", 651 | "truth_polys = [geom for geom in df_truth.geometry]\n", 652 | "truth_index = create_rtree_from_poly(truth_polys)" 653 | ] 654 | }, 655 | { 656 | "cell_type": "code", 657 | "execution_count": 105, 658 | "metadata": {}, 659 | "outputs": [ 660 | { 661 | "name": "stderr", 662 | "output_type": "stream", 663 | "text": [ 664 | "524it [00:01, 380.99it/s]\n" 665 | ] 666 | }, 667 | { 668 | "data": { 669 | "text/plain": [ 670 | "((0.7962790697674418, 428, 96, 123), (0.816793893129771, 0.7767695099818511))" 671 | ] 672 | }, 673 | "execution_count": 105, 674 | "metadata": {}, 675 | "output_type": "execute_result" 676 | } 677 | ], 678 | "source": [ 679 | "scores = evalfunction(grid_num,test_polys, truth_polys, truth_index=truth_index)\n", 680 | "scores[0],precision_recall(*scores[0][1:])" 681 | ] 682 | } 683 | ], 684 | "metadata": { 685 | "kernelspec": { 686 | "display_name": "Python [conda env:fastai-cpu]", 687 | "language": "python", 688 | "name": "conda-env-fastai-cpu-py" 689 | }, 690 | "language_info": { 691 | "codemirror_mode": { 692 | "name": "ipython", 693 | "version": 3 694 | }, 695 | "file_extension": ".py", 696 | "mimetype": "text/x-python", 697 | "name": "python", 698 | "nbconvert_exporter": "python", 699 | "pygments_lexer": "ipython3", 700 | "version": "3.6.4" 701 | } 702 | }, 703 | "nbformat": 4, 704 | "nbformat_minor": 2 705 | } 706 | -------------------------------------------------------------------------------- /archive/znz-inference-20190118.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": 1, 6 | "metadata": { 7 | "collapsed": true 8 | }, 9 | "outputs": [], 10 | "source": [ 11 | "import numpy as np\n", 12 | "\n", 13 | "import rasterio\n", 14 | "from rasterio import Affine\n", 15 | "from rasterio.windows import Window\n", 16 | "from rasterio.transform import from_bounds\n", 17 | "\n", 18 | "from pathlib import Path\n", 19 | "\n", 20 | "import matplotlib.pyplot as plt\n", 21 | "%matplotlib inline\n", 22 | "\n", 23 | "from tqdm import tqdm\n", 24 | "\n", 25 | "from skimage.transform import rescale, resize" 26 | ] 27 | }, 28 | { 29 | "cell_type": "code", 30 | "execution_count": 16, 31 | "metadata": { 32 | "collapsed": true 33 | }, 34 | "outputs": [], 35 | "source": [ 36 | "# grid_119\n", 37 | "COG_URL = 'https://oin-hotosm.s3.amazonaws.com/5ae38a540b093000130afe97/0/5ae38a540b093000130afe98.tif'" 38 | ] 39 | }, 40 | { 41 | "cell_type": "code", 42 | "execution_count": 2, 43 | "metadata": { 44 | "collapsed": true 45 | }, 46 | "outputs": [], 47 | "source": [ 48 | "# grid_029\n", 49 | "COG_URL = 'https://oin-hotosm.s3.amazonaws.com/5ae242fd0b093000130afd38/0/5ae242fd0b093000130afd39.tif'" 50 | ] 51 | }, 52 | { 53 | "cell_type": "code", 54 | "execution_count": 2, 55 | "metadata": { 56 | "collapsed": true 57 | }, 58 | "outputs": [], 59 | "source": [ 60 | "# grid_042\n", 61 | "COG_URL = 'https://oin-hotosm.s3.amazonaws.com/5ae318220b093000130afd64/0/5ae318220b093000130afd65.tif'" 62 | ] 63 | }, 64 | { 65 | "cell_type": "code", 66 | "execution_count": 3, 67 | "metadata": { 68 | "collapsed": true 69 | }, 70 | "outputs": [], 71 | "source": [ 72 | "def get_windows(rst_h, rst_w, max_h, max_w, col_off = 0, row_off = 0):\n", 73 | " wins = []\n", 74 | " rows = rst_h // max_h + 1\n", 75 | " cols = rst_w // max_w + 1 \n", 76 | " \n", 77 | " for r in range(rows):\n", 78 | " if r == rows-1: height = rst_h % max_h\n", 79 | " else: height = max_h\n", 80 | " \n", 81 | " for c in range(cols):\n", 82 | " if c == cols-1: width = rst_w % max_w\n", 83 | " else: width = max_w\n", 84 | "\n", 85 | " if width != 0 and height != 0: \n", 86 | " wins.append(((r,c),Window(c*max_w+col_off, r*max_h+row_off, width, height)))\n", 87 | " return wins" 88 | ] 89 | }, 90 | { 91 | "cell_type": "code", 92 | "execution_count": 4, 93 | "metadata": { 94 | "collapsed": true 95 | }, 96 | "outputs": [], 97 | "source": [ 98 | "def get_tfm(window, rst_full):\n", 99 | " c_o, r_o, w, h = window.flatten()\n", 100 | " left, top, right, bottom = *rst_full.xy(r_o, c_o, offset='ul'), *rst_full.xy(r_o+h, c_o+w, offset='lr')\n", 101 | " tfm = from_bounds(left,bottom,right,top, w, h)\n", 102 | " return tfm" 103 | ] 104 | }, 105 | { 106 | "cell_type": "code", 107 | "execution_count": 5, 108 | "metadata": { 109 | "collapsed": true 110 | }, 111 | "outputs": [], 112 | "source": [ 113 | "def save_subwin(arr, crs, tfm, save_fn):\n", 114 | " im = (arr*255).astype('uint8')\n", 115 | " with rasterio.open(OUTPUT/f'{save_fn}.tif', 'w', driver='GTiff', \n", 116 | " height=im.shape[0], width=im.shape[1],\n", 117 | " count=3, dtype=im.dtype, crs=crs, transform=tfm, compress='JPEG', tiled=True) as dst:\n", 118 | " for k, a in [(1, im), (2, im), (3, im)]:\n", 119 | " dst.write(a, indexes=k)" 120 | ] 121 | }, 122 | { 123 | "cell_type": "code", 124 | "execution_count": 6, 125 | "metadata": { 126 | "collapsed": true 127 | }, 128 | "outputs": [], 129 | "source": [ 130 | "def pad_window(window, pad):\n", 131 | " col_off, row_off, width, height = window.flatten()\n", 132 | " return Window(col_off-pad//2, row_off-pad//2,width+pad,height+pad)" 133 | ] 134 | }, 135 | { 136 | "cell_type": "code", 137 | "execution_count": 12, 138 | "metadata": { 139 | "collapsed": true 140 | }, 141 | "outputs": [], 142 | "source": [ 143 | "OUTPUT = Path('outputs')" 144 | ] 145 | }, 146 | { 147 | "cell_type": "code", 148 | "execution_count": 13, 149 | "metadata": {}, 150 | "outputs": [ 151 | { 152 | "data": { 153 | "text/plain": [ 154 | "{'driver': 'GTiff',\n", 155 | " 'dtype': 'uint8',\n", 156 | " 'nodata': None,\n", 157 | " 'width': 40551,\n", 158 | " 'height': 40592,\n", 159 | " 'count': 3,\n", 160 | " 'crs': CRS({'init': 'epsg:32737'}),\n", 161 | " 'transform': Affine(0.07398000359535217, 0.0, 534722.5625,\n", 162 | " 0.0, -0.07398000359535217, 9347193.0)}" 163 | ] 164 | }, 165 | "execution_count": 13, 166 | "metadata": {}, 167 | "output_type": "execute_result" 168 | } 169 | ], 170 | "source": [ 171 | "raster = rasterio.open(COG_URL,'r')\n", 172 | "raster.meta" 173 | ] 174 | }, 175 | { 176 | "cell_type": "code", 177 | "execution_count": 14, 178 | "metadata": {}, 179 | "outputs": [ 180 | { 181 | "data": { 182 | "text/plain": [ 183 | "DynamicUnet(\n", 184 | " (layers): ModuleList(\n", 185 | " (0): Sequential(\n", 186 | " (0): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n", 187 | " (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 188 | " (2): ReLU(inplace)\n", 189 | " (3): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n", 190 | " (4): Sequential(\n", 191 | " (0): BasicBlock(\n", 192 | " (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 193 | " (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 194 | " (relu): ReLU(inplace)\n", 195 | " (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 196 | " (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 197 | " )\n", 198 | " (1): BasicBlock(\n", 199 | " (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 200 | " (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 201 | " (relu): ReLU(inplace)\n", 202 | " (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 203 | " (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 204 | " )\n", 205 | " (2): BasicBlock(\n", 206 | " (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 207 | " (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 208 | " (relu): ReLU(inplace)\n", 209 | " (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 210 | " (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 211 | " )\n", 212 | " )\n", 213 | " (5): Sequential(\n", 214 | " (0): BasicBlock(\n", 215 | " (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n", 216 | " (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 217 | " (relu): ReLU(inplace)\n", 218 | " (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 219 | " (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 220 | " (downsample): Sequential(\n", 221 | " (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n", 222 | " (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 223 | " )\n", 224 | " )\n", 225 | " (1): BasicBlock(\n", 226 | " (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 227 | " (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 228 | " (relu): ReLU(inplace)\n", 229 | " (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 230 | " (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 231 | " )\n", 232 | " (2): BasicBlock(\n", 233 | " (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 234 | " (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 235 | " (relu): ReLU(inplace)\n", 236 | " (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 237 | " (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 238 | " )\n", 239 | " (3): BasicBlock(\n", 240 | " (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 241 | " (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 242 | " (relu): ReLU(inplace)\n", 243 | " (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 244 | " (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 245 | " )\n", 246 | " )\n", 247 | " (6): Sequential(\n", 248 | " (0): BasicBlock(\n", 249 | " (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n", 250 | " (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 251 | " (relu): ReLU(inplace)\n", 252 | " (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 253 | " (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 254 | " (downsample): Sequential(\n", 255 | " (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n", 256 | " (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 257 | " )\n", 258 | " )\n", 259 | " (1): BasicBlock(\n", 260 | " (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 261 | " (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 262 | " (relu): ReLU(inplace)\n", 263 | " (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 264 | " (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 265 | " )\n", 266 | " (2): BasicBlock(\n", 267 | " (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 268 | " (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 269 | " (relu): ReLU(inplace)\n", 270 | " (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 271 | " (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 272 | " )\n", 273 | " (3): BasicBlock(\n", 274 | " (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 275 | " (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 276 | " (relu): ReLU(inplace)\n", 277 | " (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 278 | " (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 279 | " )\n", 280 | " (4): BasicBlock(\n", 281 | " (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 282 | " (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 283 | " (relu): ReLU(inplace)\n", 284 | " (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 285 | " (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 286 | " )\n", 287 | " (5): BasicBlock(\n", 288 | " (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 289 | " (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 290 | " (relu): ReLU(inplace)\n", 291 | " (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 292 | " (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 293 | " )\n", 294 | " )\n", 295 | " (7): Sequential(\n", 296 | " (0): BasicBlock(\n", 297 | " (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n", 298 | " (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 299 | " (relu): ReLU(inplace)\n", 300 | " (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 301 | " (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 302 | " (downsample): Sequential(\n", 303 | " (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n", 304 | " (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 305 | " )\n", 306 | " )\n", 307 | " (1): BasicBlock(\n", 308 | " (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 309 | " (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 310 | " (relu): ReLU(inplace)\n", 311 | " (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 312 | " (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 313 | " )\n", 314 | " (2): BasicBlock(\n", 315 | " (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 316 | " (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 317 | " (relu): ReLU(inplace)\n", 318 | " (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n", 319 | " (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 320 | " )\n", 321 | " )\n", 322 | " )\n", 323 | " (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 324 | " (2): ReLU()\n", 325 | " (3): Sequential(\n", 326 | " (0): Sequential(\n", 327 | " (0): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 328 | " (1): ReLU(inplace)\n", 329 | " )\n", 330 | " (1): Sequential(\n", 331 | " (0): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 332 | " (1): ReLU(inplace)\n", 333 | " )\n", 334 | " )\n", 335 | " (4): UnetBlock(\n", 336 | " (shuf): PixelShuffle_ICNR(\n", 337 | " (conv): Sequential(\n", 338 | " (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1))\n", 339 | " )\n", 340 | " (shuf): PixelShuffle(upscale_factor=2)\n", 341 | " (pad): ReplicationPad2d((1, 0, 1, 0))\n", 342 | " (blur): AvgPool2d(kernel_size=2, stride=1, padding=0)\n", 343 | " (relu): ReLU(inplace)\n", 344 | " )\n", 345 | " (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 346 | " (conv1): Sequential(\n", 347 | " (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 348 | " (1): ReLU(inplace)\n", 349 | " )\n", 350 | " (conv2): Sequential(\n", 351 | " (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 352 | " (1): ReLU(inplace)\n", 353 | " )\n", 354 | " (relu): ReLU()\n", 355 | " )\n", 356 | " (5): UnetBlock(\n", 357 | " (shuf): PixelShuffle_ICNR(\n", 358 | " (conv): Sequential(\n", 359 | " (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1))\n", 360 | " )\n", 361 | " (shuf): PixelShuffle(upscale_factor=2)\n", 362 | " (pad): ReplicationPad2d((1, 0, 1, 0))\n", 363 | " (blur): AvgPool2d(kernel_size=2, stride=1, padding=0)\n", 364 | " (relu): ReLU(inplace)\n", 365 | " )\n", 366 | " (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 367 | " (conv1): Sequential(\n", 368 | " (0): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 369 | " (1): ReLU(inplace)\n", 370 | " )\n", 371 | " (conv2): Sequential(\n", 372 | " (0): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 373 | " (1): ReLU(inplace)\n", 374 | " )\n", 375 | " (relu): ReLU()\n", 376 | " )\n", 377 | " (6): UnetBlock(\n", 378 | " (shuf): PixelShuffle_ICNR(\n", 379 | " (conv): Sequential(\n", 380 | " (0): Conv2d(384, 768, kernel_size=(1, 1), stride=(1, 1))\n", 381 | " )\n", 382 | " (shuf): PixelShuffle(upscale_factor=2)\n", 383 | " (pad): ReplicationPad2d((1, 0, 1, 0))\n", 384 | " (blur): AvgPool2d(kernel_size=2, stride=1, padding=0)\n", 385 | " (relu): ReLU(inplace)\n", 386 | " )\n", 387 | " (bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 388 | " (conv1): Sequential(\n", 389 | " (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 390 | " (1): ReLU(inplace)\n", 391 | " )\n", 392 | " (conv2): Sequential(\n", 393 | " (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 394 | " (1): ReLU(inplace)\n", 395 | " )\n", 396 | " (relu): ReLU()\n", 397 | " )\n", 398 | " (7): UnetBlock(\n", 399 | " (shuf): PixelShuffle_ICNR(\n", 400 | " (conv): Sequential(\n", 401 | " (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1))\n", 402 | " )\n", 403 | " (shuf): PixelShuffle(upscale_factor=2)\n", 404 | " (pad): ReplicationPad2d((1, 0, 1, 0))\n", 405 | " (blur): AvgPool2d(kernel_size=2, stride=1, padding=0)\n", 406 | " (relu): ReLU(inplace)\n", 407 | " )\n", 408 | " (bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", 409 | " (conv1): Sequential(\n", 410 | " (0): Conv2d(192, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 411 | " (1): ReLU(inplace)\n", 412 | " )\n", 413 | " (conv2): Sequential(\n", 414 | " (0): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 415 | " (1): ReLU(inplace)\n", 416 | " )\n", 417 | " (relu): ReLU()\n", 418 | " )\n", 419 | " (8): PixelShuffle_ICNR(\n", 420 | " (conv): Sequential(\n", 421 | " (0): Conv2d(96, 384, kernel_size=(1, 1), stride=(1, 1))\n", 422 | " )\n", 423 | " (shuf): PixelShuffle(upscale_factor=2)\n", 424 | " (pad): ReplicationPad2d((1, 0, 1, 0))\n", 425 | " (blur): AvgPool2d(kernel_size=2, stride=1, padding=0)\n", 426 | " (relu): ReLU(inplace)\n", 427 | " )\n", 428 | " (9): MergeLayer()\n", 429 | " (10): SequentialEx(\n", 430 | " (layers): ModuleList(\n", 431 | " (0): Sequential(\n", 432 | " (0): Conv2d(99, 99, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 433 | " (1): ReLU(inplace)\n", 434 | " )\n", 435 | " (1): Sequential(\n", 436 | " (0): Conv2d(99, 99, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", 437 | " (1): ReLU(inplace)\n", 438 | " )\n", 439 | " (2): MergeLayer()\n", 440 | " )\n", 441 | " )\n", 442 | " (11): Sequential(\n", 443 | " (0): Conv2d(99, 2, kernel_size=(1, 1), stride=(1, 1))\n", 444 | " )\n", 445 | " )\n", 446 | ")" 447 | ] 448 | }, 449 | "execution_count": 14, 450 | "metadata": {}, 451 | "output_type": "execute_result" 452 | } 453 | ], 454 | "source": [ 455 | "from fastai.vision import *\n", 456 | "\n", 457 | "path = Path('data/znz-segment/znz-train-all')\n", 458 | "path_img = path/'images-512'\n", 459 | "path_lbl = path/'masks-512'\n", 460 | "get_y_fn = lambda x: path_lbl/f'{x.stem.split(\"_img\")[0]}_mask_buffered.png'\n", 461 | "codes = np.array(['Empty','Building'])\n", 462 | "unet_sz = 768\n", 463 | "bs=1\n", 464 | "\n", 465 | "data = (SegmentationItemList.from_folder(path_img)\n", 466 | " .random_split_by_pct()\n", 467 | " .label_from_func(get_y_fn, classes=codes)\n", 468 | " .transform(get_transforms(), tfm_y=True, size=unet_sz)\n", 469 | " .databunch(bs=bs)\n", 470 | " .normalize(imagenet_stats)) \n", 471 | "\n", 472 | "learn = unet_learner(data, models.resnet34)\n", 473 | "learn.load('20190108-rn34unet-comboloss-alldata-512-unfreeze-best')\n", 474 | "learn.model.eval()" 475 | ] 476 | }, 477 | { 478 | "cell_type": "code", 479 | "execution_count": 15, 480 | "metadata": { 481 | "collapsed": true 482 | }, 483 | "outputs": [], 484 | "source": [ 485 | "def get_pred(learn, tile):\n", 486 | " t_img = Image(pil2tensor(tile,np.float32))\n", 487 | " outputs = learn.predict(t_img)\n", 488 | " im = (outputs[2][1]).numpy()\n", 489 | " return im" 490 | ] 491 | }, 492 | { 493 | "cell_type": "markdown", 494 | "metadata": {}, 495 | "source": [ 496 | "# Run inference on subwindows of windows and save raster" 497 | ] 498 | }, 499 | { 500 | "cell_type": "code", 501 | "execution_count": 16, 502 | "metadata": {}, 503 | "outputs": [ 504 | { 505 | "name": "stdout", 506 | "output_type": "stream", 507 | "text": [ 508 | "512 256 768\n" 509 | ] 510 | } 511 | ], 512 | "source": [ 513 | "tif_sz = 20480 \n", 514 | "tile_sz = 512\n", 515 | "pad_sz = tile_sz//2\n", 516 | "pred_sz = tile_sz+pad_sz\n", 517 | "\n", 518 | "print(tile_sz, pad_sz, pred_sz)\n", 519 | "assert pred_sz == unet_sz\n", 520 | "\n", 521 | "scale_factor = 1\n", 522 | "save_prefix = f'grid_042_20190118_test{scale_factor}x'" 523 | ] 524 | }, 525 | { 526 | "cell_type": "code", 527 | "execution_count": null, 528 | "metadata": { 529 | "collapsed": true, 530 | "scrolled": true 531 | }, 532 | "outputs": [], 533 | "source": [ 534 | "tile_scaled = tile_sz*scale_factor\n", 535 | "pad_scaled = pad_sz*scale_factor\n", 536 | "test_wins = get_windows(raster.meta['height'], raster.meta['width'], tif_sz, tif_sz)\n", 537 | "\n", 538 | "print(len(test_wins))\n", 539 | "for idx, win in enumerate(test_wins):\n", 540 | " print(idx, win)\n", 541 | " \n", 542 | " # make subwindows and blank array to fill in\n", 543 | " col_off, row_off, rst_w, rst_h = win[1].flatten()\n", 544 | " sub_wins = get_windows(rst_h, rst_w, tile_scaled, tile_scaled, col_off, row_off)\n", 545 | " new_arr = np.zeros((rst_h, rst_w))\n", 546 | "\n", 547 | " for (row_idx, col_idx), window in tqdm(sub_wins):\n", 548 | " win_padded = pad_window(window, pad_scaled)\n", 549 | " win_img = np.rollaxis(raster.read(window=win_padded, boundless=True),0,3)/255\n", 550 | "\n", 551 | " # scale down windowed read to unet input size\n", 552 | " win_img = rescale(win_img,1/scale_factor,anti_aliasing=False)\n", 553 | " placeholder = np.zeros((pred_sz,pred_sz,3))\n", 554 | " placeholder[:win_img.shape[0],:win_img.shape[1]] = win_img\n", 555 | " \n", 556 | " # skip inference if empty window\n", 557 | " if placeholder.max() > 0: pr = get_pred(learn, placeholder)\n", 558 | " else: pr = placeholder[:,:,0]\n", 559 | " \n", 560 | " # scale back up to original tile size to fill into blank array at right place\n", 561 | " pr = rescale(pr, scale_factor, anti_aliasing=False)\n", 562 | " pr = pr[pad_scaled//2:-pad_scaled//2,pad_scaled//2:-pad_scaled//2]\n", 563 | "\n", 564 | " try: \n", 565 | " width, height = window.flatten()[-2:]\n", 566 | " start_y = row_idx*tile_scaled\n", 567 | " start_x = col_idx*tile_scaled\n", 568 | " new_arr[start_y:start_y+height, start_x:start_x+width]= pr[:height,:width]\n", 569 | " except Exception as exc: print(f'{exc}')\n", 570 | " \n", 571 | " tfm = get_tfm(win[1], raster)\n", 572 | " save_subwin(new_arr, raster.meta['crs'].data['init'], tfm, f'{save_prefix}_id{idx}')" 573 | ] 574 | }, 575 | { 576 | "cell_type": "code", 577 | "execution_count": 52, 578 | "metadata": {}, 579 | "outputs": [ 580 | { 581 | "data": { 582 | "text/plain": [ 583 | "" 584 | ] 585 | }, 586 | "execution_count": 52, 587 | "metadata": {}, 588 | "output_type": "execute_result" 589 | }, 590 | { 591 | "data": { 592 | "image/png": "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\n", 593 | "text/plain": [ 594 | "
" 595 | ] 596 | }, 597 | "metadata": { 598 | "needs_background": "light" 599 | }, 600 | "output_type": "display_data" 601 | } 602 | ], 603 | "source": [ 604 | "plt.figure(figsize=(15,15))\n", 605 | "plt.imshow(new_arr)" 606 | ] 607 | }, 608 | { 609 | "cell_type": "markdown", 610 | "metadata": {}, 611 | "source": [ 612 | "## Merge rasters" 613 | ] 614 | }, 615 | { 616 | "cell_type": "code", 617 | "execution_count": 18, 618 | "metadata": { 619 | "collapsed": true 620 | }, 621 | "outputs": [], 622 | "source": [ 623 | "from rasterio.merge import merge" 624 | ] 625 | }, 626 | { 627 | "cell_type": "code", 628 | "execution_count": 19, 629 | "metadata": { 630 | "collapsed": true 631 | }, 632 | "outputs": [], 633 | "source": [ 634 | "fns = sorted([o.name for o in OUTPUT.iterdir() if save_prefix in o.name])" 635 | ] 636 | }, 637 | { 638 | "cell_type": "code", 639 | "execution_count": 20, 640 | "metadata": {}, 641 | "outputs": [ 642 | { 643 | "data": { 644 | "text/plain": [ 645 | "['grid_042_20190118_test1x_id0.tif',\n", 646 | " 'grid_042_20190118_test1x_id1.tif',\n", 647 | " 'grid_042_20190118_test1x_id2.tif',\n", 648 | " 'grid_042_20190118_test1x_id3.tif']" 649 | ] 650 | }, 651 | "execution_count": 20, 652 | "metadata": {}, 653 | "output_type": "execute_result" 654 | } 655 | ], 656 | "source": [ 657 | "fns" 658 | ] 659 | }, 660 | { 661 | "cell_type": "code", 662 | "execution_count": 21, 663 | "metadata": { 664 | "collapsed": true 665 | }, 666 | "outputs": [], 667 | "source": [ 668 | "src_files_to_mosaic = []\n", 669 | "\n", 670 | "for fn in fns:\n", 671 | " src = rasterio.open(OUTPUT/fn)\n", 672 | " src_files_to_mosaic.append(src)" 673 | ] 674 | }, 675 | { 676 | "cell_type": "code", 677 | "execution_count": 22, 678 | "metadata": { 679 | "collapsed": true 680 | }, 681 | "outputs": [], 682 | "source": [ 683 | "mosaic, out_tfm = merge(src_files_to_mosaic)" 684 | ] 685 | }, 686 | { 687 | "cell_type": "code", 688 | "execution_count": 23, 689 | "metadata": {}, 690 | "outputs": [ 691 | { 692 | "data": { 693 | "text/plain": [ 694 | "(Affine(0.07398361590021522, 0.0, 534722.5625,\n", 695 | " 0.0, -0.07398361590021522, 9347193.0), (3, 40592, 40551))" 696 | ] 697 | }, 698 | "execution_count": 23, 699 | "metadata": {}, 700 | "output_type": "execute_result" 701 | } 702 | ], 703 | "source": [ 704 | "out_tfm, mosaic.shape" 705 | ] 706 | }, 707 | { 708 | "cell_type": "code", 709 | "execution_count": 24, 710 | "metadata": {}, 711 | "outputs": [ 712 | { 713 | "data": { 714 | "text/plain": [ 715 | "{'driver': 'GTiff',\n", 716 | " 'dtype': 'uint8',\n", 717 | " 'nodata': None,\n", 718 | " 'width': 20071,\n", 719 | " 'height': 20112,\n", 720 | " 'count': 3,\n", 721 | " 'crs': CRS({'init': 'epsg:32737'}),\n", 722 | " 'transform': Affine(0.07398368951053305, 0.0, 536237.6729736328,\n", 723 | " 0.0, -0.0739836819964856, 9345677.889526367)}" 724 | ] 725 | }, 726 | "execution_count": 24, 727 | "metadata": {}, 728 | "output_type": "execute_result" 729 | } 730 | ], 731 | "source": [ 732 | "out_meta = src.meta.copy()\n", 733 | "out_meta" 734 | ] 735 | }, 736 | { 737 | "cell_type": "code", 738 | "execution_count": 25, 739 | "metadata": { 740 | "collapsed": true 741 | }, 742 | "outputs": [], 743 | "source": [ 744 | "out_meta.update({\"driver\": \"GTiff\",\n", 745 | " \"count:\":1,\n", 746 | " \"height\": mosaic.shape[1],\n", 747 | " \"width\": mosaic.shape[2],\n", 748 | " \"transform\": out_tfm,\n", 749 | " \"compress\": \"jpeg\",\n", 750 | " \"tiled\": True\n", 751 | " })" 752 | ] 753 | }, 754 | { 755 | "cell_type": "code", 756 | "execution_count": 26, 757 | "metadata": { 758 | "collapsed": true 759 | }, 760 | "outputs": [], 761 | "source": [ 762 | "with rasterio.open(OUTPUT/f'{save_prefix}_merged.tif', \"w\", **out_meta) as dest:\n", 763 | " dest.write(mosaic)" 764 | ] 765 | } 766 | ], 767 | "metadata": { 768 | "kernelspec": { 769 | "display_name": "Python [default]", 770 | "language": "python", 771 | "name": "python3" 772 | }, 773 | "language_info": { 774 | "codemirror_mode": { 775 | "name": "ipython", 776 | "version": 3 777 | }, 778 | "file_extension": ".py", 779 | "mimetype": "text/x-python", 780 | "name": "python", 781 | "nbconvert_exporter": "python", 782 | "pygments_lexer": "ipython3", 783 | "version": "3.5.3" 784 | } 785 | }, 786 | "nbformat": 4, 787 | "nbformat_minor": 2 788 | } 789 | -------------------------------------------------------------------------------- /static/grid119_preview.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/daveluo/zanzibar-aerial-mapping/5e78ff82b0fadcd809946c6756562553ab2f24ac/static/grid119_preview.png -------------------------------------------------------------------------------- /static/overview_predict.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/daveluo/zanzibar-aerial-mapping/5e78ff82b0fadcd809946c6756562553ab2f24ac/static/overview_predict.png -------------------------------------------------------------------------------- /static/overview_train.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/daveluo/zanzibar-aerial-mapping/5e78ff82b0fadcd809946c6756562553ab2f24ac/static/overview_train.png -------------------------------------------------------------------------------- /static/znz-demo.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/daveluo/zanzibar-aerial-mapping/5e78ff82b0fadcd809946c6756562553ab2f24ac/static/znz-demo.gif --------------------------------------------------------------------------------