Research map / 3D vision / Depth and stereo
Wat3R: Underwater 3D Geometry Learning without Underwater Annotations
Jiangwei Ren, Xingyu Jiang, Zijie Song, Wei Xu, Hongkai Lin, Dingkang Liang, Xiang Bai
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 |
|---|---|---|---|
| adapt VGGT with Wat3R semi-supervised training | VGGT · Sea-thru multi-view depth | Rel: 0.190 -> 0.167 | Improved |
| adapt VGGT with Wat3R semi-supervised training | VGGT · FLSea Stereo multi-view depth | Rel: 0.137 -> 0.119 | Improved |
| adapt VGGT with Wat3R | VGGT · Water3D point map | Overall mean: 0.624 -> 0.427 | Improved |
| adapt VGGT with Wat3R | VGGT · SeaThru-NeRF pose | AUC@5: 0.392 -> 0.540 | Improved |
| adapt VGGT with Wat3R | VGGT · SeaThru-NeRF pose | AUC@5: 0.540 vs 0.731 (DA3) | Hurt |
| adapt VGGT with Wat3R | VGGT · FLSea VI monocular depth | Rel: 0.107 -> 0.061 | Improved |
| apply underwater image enhancement before VGGT (Semi-UIR) | VGGT · Sea-thru multi-view depth | Rel: 0.190 -> 0.201 | Hurt |
| apply underwater image enhancement before VGGT (PSPL) | VGGT · FLSea Stereo multi-view depth | Rel: 0.137 -> 0.160 | Hurt |
| add synthetic underwater rendering training (ablation) | VGGT · Sea-thru multi-view depth | Rel: 0.190 -> 0.173 | Improved |
| add real unlabeled underwater video to synthetic training (ablation) | Wat3R ablation · Sea-thru multi-view depth | Rel: 0.173 -> 0.181 | Hurt |
| add real unlabeled underwater video to synthetic training (ablation) | Wat3R ablation · FLSea Stereo multi-view depth | Rel: 0.135 -> 0.165 | Hurt |
| add strong augmentation with real video (ablation) | Wat3R ablation · FLSea Stereo multi-view depth | Rel: 0.165 -> 0.126 | Improved |
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.
- VGGT — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 10 of 10 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
- AquaSplatting: A Hybrid 3D Representation for Robust Underwater Scene Reconstruction via Dual-Branch Rendering (AAAI 2026 · similarity 0.89)
- Plenodium: Underwater 3D Scene Reconstruction with Plenoptic Medium Representation (NeurIPS 2025 · similarity 0.87)
- Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion (CVPR 2024 · similarity 0.87)
- Fin3R: Fine-tuning Feed-forward 3D Reconstruction Models via Monocular Knowledge Distillation (NeurIPS 2025 · similarity 0.86)
- NAUTILUS: A Large Multimodal Model for Underwater Scene Understanding (NeurIPS 2025 · similarity 0.85)
- A Unified Image-Dense Annotation Generation Model for Underwater Scenes (CVPR 2025 · similarity 0.85)
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.
