Microsoft open-sources Anaximander to run any geospatial deep learning model inside QGIS
anselm · x · 2026-11-03
Microsoft's AI for Good Research Lab released Anaximander (arXiv:2610.09085), an open-source system that lowers the friction of applying deep learning to satellite imagery.
- The problem: Pointing GPT-image-1, Meta's SAM 3 and a specialist field-mapping model at the same farmland, the general-purpose models produced the least reliable maps. Every model ships in its own format targeting different compute backends, so analysts must hand-write model loading, image tiling, per-tile inference, stitching and georeferencing code — discouraging systematic comparison, and teams settle for whichever model they got running first.
- The solution: Anaximander unifies model source and compute location behind one interactive interface. An inference server loads models from multiple common sources, serves them on any backend from local workstations to serverless cloud, and provides session management, model caching, and per-tile streaming results.
- Integration: A QGIS plugin drives tiling, result reassembly, georeferencing and real-time per-tile status visualization. Use cases include field delineation, crop monitoring and disaster response.
More from Research
- NVIDIA's AgentBaton cuts agent API costs 65.9% via well-timed model handoffs — amaarora · 2026-11-03
- SJTU-led paper proposes 'Avatar Intelligence' and an Agentic Oversight Index for personal agents — SJTU · 2026-11-03
- Vapor-cell atomic E-field sensor hits 0.2 mV/m/√Hz in 11 mm³, no electrodes — jwt0625 · 2026-11-03
- Anthropic Open-Sources FlashProve, Zero-Knowledge Proofs of AI Inference, Tested on Llama 4 Maverick — yenkel · 2026-11-03
- BiGym 2.0: Claude-written policies hit 51.9% on humanoid household tasks, nearing VLA — teortaxesTex · 2026-11-03
- Hopfield model thermodynamic limit proven at all loads without replica-symmetry assumptions — _onionesque · 2026-11-03