What problem does it solve? Answering spatial questions about images and video—distance, size, orientation, relative position, camera motion, and cross-frame object counts—requires structured geometry rather than raw visual inspection. This Skill builds a coarse scene from grounded object boxes and depth estimates, then applies stateless geometry tools to reason about layout, viewpoints, and motion. ## Core Features & Use Cases - Scene Construction: Combine per-frame object boxes, depth estimates, and camera motions into a shared world-coordinate scene with bird's-eye-view visualization. - Spatial Analysis Tools: Triangulate distances, calibrate scale from known object sizes, track object and camera motion, count and match entities across frames, and reason about visibility from arbitrary viewpoints. - Exploration Planning: Assess coverage and reachability, then plan navigation or search steps for targets that are unobserved or grounded. - Use Case: Given a multi-room video walkthrough, ground furniture in sampled frames, build a scene, and determine whether a sofa is to the left or right of a table from a virtual viewpoint at the door facing the window. ## Quick Start Ask the agent to determine how far the chair is from the desk in this video and whether the camera moved between frames.
