Research map / 3D vision / Depth and stereo
WildRayZer: Self-supervised Large View Synthesis in Dynamic Environments
Xuweiyi Chen, Wentao Zhou, Zezhou Cheng
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 |
|---|---|---|---|
| Learned motion masking with masked renderer vs RayZer + off-the-shelf SAV masks | RayZer · D-RE10K-Mask (2 views) | PSNR_s: 19.01 -> 21.78 | Improved |
| WildRayZer vs WildGaussians on sparse-view NVS | WildGaussians · D-RE10K-iPhone (4 views) | PSNR: 20.44 -> 20.98 | Improved |
| Naive token masking with accurate off-the-shelf masks | RayZer · Qualitative | blur | Hurt |
| Add learned masked renderer to feed-forward baseline | RayZer + SAV · D-RE10K-iPhone transient regions (2 views) | PSNR_t: 17.77 -> 20.99 | Improved |
| WildRayZer motion estimator vs Segment Any Motion | Motion segmentation · D-RE10K-Mask (n=2) | mIoU: 31.9 -> 53.9 | Improved |
| Add copy-paste augmentation to pseudo-mask training (ablation) | Motion estimator · D-RE10K | mIoU: 53.9 -> 53.9 | No clear effect |
| Add copy-paste augmentation to pseudo-mask training (ablation) | Motion estimator · D-RE10K-iPhone | mIoU: 45.3 -> 49.7 | Improved |
| Add copy-paste augmentation (ablation) | Motion estimator · DAVIS (8 sequences) | mIoU: 3.4 -> 31.0 | Improved |
| Use copy-paste alone without pseudo-masks (ablation) | Motion estimator · D-RE10K | mIoU: 18.2 vs 53.9 | Hurt |
| Add DINOv3 features to motion estimator (ablation) | Motion estimator · D-RE10K | mIoU: 29.4 -> 39.4 | Improved |
| Add DINOv3 features to motion estimator (ablation) | Motion estimator · training speed | steps to mIoU=30: 20k -> 1.5k | Improved |
Later papers that compared with this method
1 later accepted paper reports WildRayZer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Dynamic-Robust Photometric–Semantic Reconstruction for Open-Vocabulary 3D Scene Understanding (ECCV 2026 · better on 7 of 9 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- WildGaussians — WildGaussians: 3D Gaussian Splatting In the Wild (NeurIPS 2024) · better on 5 of 5 shared results
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
- WildCAT3D: Appearance-Aware Multi-View Diffusion in the Wild (NeurIPS 2025 · similarity 0.87)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.87)
- Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videos (NeurIPS 2025 · similarity 0.87)
- NViST: In the Wild New View Synthesis from a Single Image with Transformers (CVPR 2024 · similarity 0.87)
- Dynamic View Synthesis as an Inverse Problem (NeurIPS 2025 · similarity 0.86)
- One View Is Enough: In-the-Wild Monocular Pretraining for Novel View Generation (NeurIPS 2026 · 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.
