Meta's Reality Labs Research has published ShapeR, an open research project that reconstructs metric-accurate 3D scenes from casually captured image sequences, and has released code, models and an evaluation dataset for developers to try. The system takes ordinary multi-view captures — the kind produced by a smartphone or AR glasses — and generates object-centric 3D meshes that preserve real-world scale, addressing a persistent weakness in earlier single-image approaches.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.