![]() SpaceNet 7 has demonstrated that identifying, tracking, and detecting change in precise building footprints using moderate resolution (4m) Planet imagery is possible. Quantifying population statistics is fundamental to dozens of the United Nations Sustainable Development Goals, yet many countries lack effective civil registration systems. The winning algorithms are available under a permissive license at: Īll algorithms leverage Docker for easy deployment, and be either trained on new data or tested using existing model weights. Tracking and Change Illustrationįigure 3 below provides an example of the winning model’s ability to correctly track building identifiers over time, as well as catalog change. Performance in various cities for the winning SpaceNet 7 model. Table 1 below details the specifics of each model, while Figure 1 illustrates the SpaceNet Change and Object Tracking ( SCOT) score vs inference speed (for an in-depth analysis of SCOT scores, see our previous blog).įigure 2. Recall from our results announcement blog that the top-5 submissions to SpaceNet improved markedly upon the SpaceNet 7 algorithmic baseline, and applied a variety of clever approaches to improve performance. In the sections below we explore performance with an interactive map, and provide details on how to access the models. Furthermore, our paper on the SpaceNet 7 dataset and initial results can now be found on arXiv. ![]() This post serve as the final installment in the SpaceNet 7 analysis series, and we are pleased to announce that the winning models are now available in the open source. The SpaceNet 7 Multi-Temporal Urban Development Challenge yielded some impressive results, with some interesting phenomena noted in analysis blogs 1, 2. ![]() īuilding tracking and change detection from moderate satellite imagery is surprisingly tractable, as we’ve demonstrated in the last few DownLinQ blogs. SpaceNet is run in collaboration by co-founder and managing partner CosmiQ Works, co-founder and co-chair Maxar Technologies, and our partners including Amazon Web Services (AWS), Capella Space, Topcoder, IEEE GRSS, the National Geospatial-Intelligence Agency and Planet. Preface: SpaceNet LLC is a nonprofit organization dedicated to accelerating open source, artificial intelligence applied research for geospatial applications, specifically foundational mapping (i.e., building footprint & road network detection).
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |