![](\"https://img.shields.io/badge/Follow-%40RoboSatPink-ff69b4.svg\"){kind=icon}
![](\"https://img.shields.io/gitter/room/robosatpink/community.svg?color=ff69b4&style=popout\")
\n ![\"RoboSat.pink](\"https://pbs.twimg.com/media/DpjonykWwAANpPr.jpg\")
\n
![\"Draw](\"https://raw.githubusercontent.com/datapink/robosat.pink/master/docs/img/readme/draw_me_robosat_pink.png\")
\n\n\n \nDocumentation:\n--------------\n\n### Tutorials:\n- RoboSat.pink 101\n- How to use plain OpenData to create a decent training OpenDataSet\n- How to extend RoboSat.pink features, and use it as a Framework\n\n### Config file:\n- RoboSat.pink configuration file\n\n### Tools:\n\n- `rsp cover` Generate a tiles covering, in csv format: X,Y,Z\n- `rsp download` Downloads tiles from a remote server (XYZ, WMS, or TMS)\n- `rsp extract` Extracts GeoJSON features from OpenStreetMap .pbf\n- `rsp rasterize` Rasterize vector features (GeoJSON or PostGIS), to raster tiles\n- `rsp subset` Filter images in a slippy map dir using a csv tiles cover\n- `rsp tile` Tile raster coverage\n- `rsp train` Trains a model on a dataset\n- `rsp export` Export a model to ONNX or Torch JIT\n- `rsp predict` Predict masks, from given inputs and an already trained model\n- `rsp compare` Compute composite images and/or metrics to compare several XYZ dirs\n- `rsp vectorize` Extract simplified GeoJSON features from segmentation masks\n- `rsp info` Print RoboSat.pink version informations\n\n### Presentations slides:\n - @SOTM 2019\n - @FOSS4G 2019\n\n\n\n\n\n\nInstalls:\n--------\n\n### With PIP:\n```\npip3 install RoboSat.pink # For latest stable version\n```\n\nor\n\n```\npip3 install git+https://github.com/datapink/robosat.pink # For current dev version\n```\n\n### With Ubuntu 19.04, from scratch:\n\n```\nsudo sh -c \"apt update && apt install -y build-essential python3-pip\"\npip3 install RoboSat.pink && export PATH=$PATH:~/.local/bin\nwget http://us.download.nvidia.com/XFree86/Linux-x86_64/430.40/NVIDIA-Linux-x86_64-430.40.run\nsudo sh NVIDIA-Linux-x86_64-430.40.run -a -q --ui=none\n```\n\n### With CentOS 7, from scratch:\n```\nsudo sh -c \"yum -y update && yum install -y python36 wget && python3.6 -m ensurepip\"\npip3 install --user RoboSat.pink\nsudo sh -c \"yum groupinstall -y 'Development Tools' && yum install -y kernel-devel epel-release\"\nwget http://us.download.nvidia.com/XFree86/Linux-x86_64/430.40/NVIDIA-Linux-x86_64-430.40.run\nsudo sh NVIDIA-Linux-x86_64-430.40.run -a -q --ui=none\n```\n\n\n### NOTAS: \n- Requires: Python 3.6 or 3.7\n- GPU is not strictly mandatory, but `rsp train` and `rsp predict` would be -that- slower without.\n- To test RoboSat.pink install, launch in a terminal: `rsp -h` or `rsp info`\n- Upon your ```pip``` PATH setting, you may have to update it: ```export PATH=$PATH:.local/bin```\n- If needed, to remove pre-existing Nouveau driver: ```sudo sh -c \"echo blacklist nouveau > /etc/modprobe.d/blacklist-nvidia-nouveau.conf && update-initramfs -u && reboot\"```\n\n\n\n\nArchitecture:\n------------\n\nRoboSat.pink use cherry-picked Open Source libs among Deep Learning, Computer Vision and GIS stacks.\n\n![\"Stacks\"](\"https://raw.githubusercontent.com/datapink/robosat.pink/master/docs/img/readme/stacks.png\")
\n\n\n\nGeoSpatial OpenDataSets:\n------------------------\n- Christoph Rieke's Awesome Satellite Imagery Datasets\n- Zhang Bin, Earth Observation OpenDataset blog \n\n\nBibliography:\n-------------\n\n- U-Net: Convolutional Networks for Biomedical Image Segmentation\n- Deep Residual Learning for Image Recognition\n- Angiodysplasia Detection and Localization Using Deep\nConvolutional Neural Networks\n- TernausNetV2: Fully Convolutional Network for Instance Segmentation\n- The Lov\u00e1sz-Softmax loss: A tractable surrogate for the optimization of the IoU measure in neural networks\n- Albumentations: fast and flexible image augmentations\n\n\n\n\n\n\n\n\n\n\nContributions and Services:\n---------------------------\n\n- Pull Requests are welcome ! Feel free to send code...\n Don't hesitate either to initiate a prior discussion via gitter or ticket on any implementation question.\n And give also a look at Makefile rules.\n\n- If you want to collaborate through code production and maintenance on a long term basis, please get in touch, co-edition with an ad hoc governance can be considered.\n\n- If you want a new feature, but don't want to implement it, DataPink provide core-dev services.\n\n- Expertise and training on RoboSat.pink are also provided by DataPink.\n\n- And if you want to support the whole project, because it means for your own business, funding is also welcome.\n\n\n\n\n### Requests for funding:\n\n- Increase (again) prediction accuracy :\n - on low resolution imagery\n - even with few labels\n - feature extraction when they are (really) close\n - with multibands and Data Fusion\n\n- Add support for\u00a0:\n - MultiClass classification\n - Linear features extraction\n - PointCloud data support\n - Time Series Imagery\n - StreetView Imagery\n \n- Improve (again) performances\n\n\n\n\nAuthors:\n--------\n- Olivier Courtin