Reconstructing 3D from drone footage using gimbal telemetry and CV tricks, with agents confirming object details
doodlestein · x · 2026-09-08
The author explains how to extract 3D structure from drone video when the camera is essentially static and only the gimbal sweeps: use telemetry at 25fps (matching every other video frame) with computer vision registration tricks.
A clever agentic layer: when the agent spots a car, it asks you to confirm make, model, and year so it can look up exact dimensions for absolute positioning; it also looks for reference objects like Starlink dishes.
Related event: Rebuilding House 3D Scenes from a $500 Drone's Video(3 posts)→
More from Research
- Meta FAIR's Self-Direction turns agent failures into reusable LoRA skills, +13.8pp on Terminal-Bench — karen_ullrich · 2026-10-27
- DONIC hits 86.79% on bimanual dexterous tracking, crushing prior SOTA's 9.43% — liuziwei7 · 2026-10-27
- Why don't more LLMs ship 10M-token context? A researcher asks about attention tradeoffs — agarwl_ · 2026-10-27
- PSNR below 15 dB in papers likely means the metric was misused, researcher warns — kwangmoo_yi · 2026-10-27
- Ortet's first health eval: 34 open-weight models, 50 tasks, generic rankings fall short — thegautamkamath · 2026-10-27
- Scaling-fit discrepancy: output-head FLOPs are ~65% of compute at the smallest rung, ~3% at 535B — dlwh · 2026-10-27