[{"content":"Getting Selected I\u0026rsquo;m excited to announce that I\u0026rsquo;ve been selected for Google Summer of Code 2026 with NumFOCUS, working on the DeepForest project!\nMy project, \u0026ldquo;Recovering Historical Image Data Using Automated OrthoRegistration,\u0026rdquo; focuses on building an automated pipeline to align historical forest health survey annotations with modern satellite imagery.\nThe Problem The U.S. Forest Service\u0026rsquo;s Aerial Detection Survey (ADS) program has produced tens of thousands of hand-drawn polygons marking forest damage from aircraft. Due to GPS imprecision and flight geometry, these polygons are spatially offset by 50-500 meters from their true locations.\nThis means we can\u0026rsquo;t directly use these annotations as training data for modern detection models — the labels don\u0026rsquo;t line up with what\u0026rsquo;s actually visible in satellite imagery.\nMy Approach I\u0026rsquo;m building a hybrid alignment pipeline that combines:\nGeometric transformations — affine and projective corrections based on known control points Learned displacement estimation — a self-supervised deep learning model that predicts per-polygon correction vectors Energy-function optimization — combining spectral similarity, edge alignment, and spatial coherence to evaluate registration quality What\u0026rsquo;s Next Over the coming weeks, I\u0026rsquo;ll be:\nSetting up the development environment and understanding the existing DeepForest codebase Prototyping the initial alignment pipeline Starting work on the learned refinement layer Stay tuned for updates!\nThis is part of my GSoC 2026 blog series. Follow along with the #gsoc tag.\n","permalink":"https://musaqlain.dev/blog/gsoc-2026-community-bonding/","summary":"\u003ch2 id=\"getting-selected\"\u003eGetting Selected\u003c/h2\u003e\n\u003cp\u003eI\u0026rsquo;m excited to announce that I\u0026rsquo;ve been selected for \u003cstrong\u003eGoogle Summer of Code 2026\u003c/strong\u003e with \u003cstrong\u003eNumFOCUS\u003c/strong\u003e, working on the \u003ca href=\"https://github.com/weecology/DeepForest\"\u003eDeepForest\u003c/a\u003e project!\u003c/p\u003e\n\u003cp\u003eMy project, \u003cem\u003e\u0026ldquo;Recovering Historical Image Data Using Automated OrthoRegistration,\u0026rdquo;\u003c/em\u003e focuses on building an automated pipeline to align historical forest health survey annotations with modern satellite imagery.\u003c/p\u003e\n\u003ch2 id=\"the-problem\"\u003eThe Problem\u003c/h2\u003e\n\u003cp\u003eThe U.S. Forest Service\u0026rsquo;s Aerial Detection Survey (ADS) program has produced tens of thousands of hand-drawn polygons marking forest damage from aircraft. Due to GPS imprecision and flight geometry, these polygons are spatially offset by 50-500 meters from their true locations.\u003c/p\u003e","title":"GSoC 2026: Community Bonding Period"},{"content":"","permalink":"https://musaqlain.dev/projects/","summary":"","title":"Projects"},{"content":"","permalink":"https://musaqlain.dev/research/","summary":"","title":"Research"}]