CompVis open-sources RayDer: self-supervised novel view synthesis from real video with LLM-like power-law scaling
rsasaki0109 · x · 2026-11-10
CompVis at LMU Munich (Björn Ommer's group, ECCV 2026) presents RayDer, which removes the static-scene data bottleneck of self-supervised novel view synthesis by training stably on general dynamic real-world video. It consolidates three networks into one unified transformer, adds dynamic state prediction with dropout, and improves pose learning via autoregressive training.
Zero-shot comparisons against E-RayZer across DTU, CO3D, Mip360 and LLFF settings — including near-zero context-view overlap — show clearly sharper renders. Performance scales with data, model size and compute following power laws (R² > 0.99), analogous to LLM scaling. A minimal PyTorch re-implementation is open-sourced on GitHub.
Related event: CompVis Open-Sources RayDer for Self-Supervised Novel View Synthesis(2 posts)→
More from Research
- Four Free Math Books for ML: Linear Algebra, Probability, Optimization, Information Theory — cneuralnetwork · 2026-11-10
- 4 free math books that actually matter for ML: linear algebra to information theory — cneuralnetwork · 2026-11-10
- Meta FAIR's e-means beats k-means in 1/10 epochs with 10-45x GPU speedups — burny_tech · 2026-11-10
- NeurIPS workshop paper: gradient-based data attribution can't reliably filter subliminal learning — ponguru · 2026-11-10
- Yang-Mills Working Group to Release OpenAI's Full Lean Code and Validation Tomorrow — burny_tech · 2026-11-10
- Interpretability researchers reminded: explanation matters, not just predictive power — burny_tech · 2026-11-10