Research map / 3D vision / Depth and stereo
RayDer: Scalable Self-Supervised Novel View Synthesis from Real-World Video
Ulrich Prestel, Stefan Andreas Baumann, Nick Stracke, Bjorn Ommer
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add dynamic state prediction to the RayZer-like baseline (ablation) | RayDer-S (SA-V) · RE10K NVS | PSNR: 22.53 -> 24.01 | Improved |
| add dynamic state prediction (camera estimation side effect) (ablation) | RayDer-S (SA-V) · DL3DV-10k camera estimation | R@10: 59.8 -> 56.1 | Hurt |
| add state dropout on top of dynamic state (ablation) | RayDer-S (SA-V) · DL3DV-10k camera estimation | R@10: 56.1 -> 62.4 | Improved |
| add state dropout (inference without state) (ablation) | RayDer-S (SA-V) · RE10K NVS without state | PSNR: 13.42 -> 23.01 | Improved |
| add single-network consolidation of the three networks (ablation) | RayDer-S (SA-V) · RE10K NVS | PSNR: 23.76 -> 25.33 | Improved |
| add single-network consolidation (ablation) | RayDer-S (SA-V) · DL3DV-10k camera estimation | R@10: 62.4 -> 68.8 | Improved |
| add parallel-target attention (ablation) | RayDer-S (SA-V) · RE10K NVS | PSNR: 25.33 -> 25.12 | Mixed |
| add ordered autoregression over views (ablation) | RayDer-S (SA-V) · RE10K NVS / DL3DV camera | PSNR / R@10: PSNR 25.12 -> 24.49; R@10 70.1 -> 73.6 | Mixed |
| use random-order autoregression instead of no autoregression (ablation) | RayDer-S (SA-V) · RE10K NVS | PSNR: 25.12 -> 26.28 | Improved |
| add random-order autoregression (ablation) | RayDer-S (SA-V) · DL3DV-10k camera estimation | R@10: 70.1 -> 84.4 | Improved |
| add local high-resolution layers (ablation) | RayDer-S (SA-V) · RE10K NVS | PSNR: 26.28 -> 26.87 | Improved |
| add random-order autoregression (SV-HQ training) (ablation) | RayDer-S (SV-HQ) · RE10K NVS | PSNR: 26.21 -> 29.57 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- RayZer — RayZer: A Self-supervised Large View Synthesis Model (ICCV 2025) · better on 2 of 2 shared results
- GS-LRM — GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RayZer: A Self-supervised Large View Synthesis Model (ICCV 2025 · similarity 0.92)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.86)
- NViST: In the Wild New View Synthesis from a Single Image with Transformers (CVPR 2024 · similarity 0.86)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.86)
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.86)
- Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videos (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
