Skip to Main Content

ArborPulse

Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

ArborPulse — Demo video

ArborPulse turns short videos of trees in wind into traceable inspection evidence for arborists and field crews. It addresses the gap between noticing movement and collecting a recording that supports useful professional review. The workflow combines two Muse Spark 1.3 steps with a deterministic motion engine. With consent, Meta reviews six timestamped stills, field notes, and the operator's tree/reference selections. It proposes tracking patches and flags capture limitations. The operator confirms both regions, then PhysicsLens measures projected motion locally in the browser and compensates for camera translation using the fixed reference. Muse Spark combines the resulting measurements, source frames, and recorder notes into a brief with evidence links and questions for site inspection. Users can export measurements and reports; the full video stays on the device. The demo shows licensed real footage receiving a recapture recommendation when tracking is unreliable. A labelled synthetic control with a 0.75 Hz input produces about 0.78 Hz, within the available frequency resolution. This tests the software without claiming accuracy on real trees. Built during the hackathon, ArborPulse helps prepare repeatable captures and reviewable inspection records. It does not diagnose hidden decay, prescribe pruning, or certify tree safety. The next step is field evaluation with arborists and independent measurements.

Team