You can now calculate the exact 3D orientation of an object just by looking at its flat shadow.
April 15, 2026
Original Paper
Globally Optimal Pose from Orthographic Silhouettes
arXiv · 2604.09199
The Takeaway
Estimating 3D pose usually requires identifying specific points on an object, which is impossible if the object is complex or has holes. This method is the first to efficiently find the globally optimal 3D pose using only 2D silhouettes—no point-to-point correspondence needed. This works even for 'holey' objects where standard algorithms fail. It’s a massive jump for automated assembly and spatial computing where texture and color aren't reliable. It makes 'shadow-based' tracking a viable, high-precision tool for the first time.
From the abstract
We solve the problem of determining the pose of known shapes in $\mathbb{R}^3$ from their unoccluded silhouettes. The pose is determined up to global optimality using a simple yet under-explored property of the area-of-silhouette: its continuity w.r.t trajectories in the rotation space. The proposed method utilises pre-computed silhouette-signatures, modelled as a response surface of the area-of-silhouettes. Querying this silhouette-signature response surface for pose estimation leads to a stron