Research map / 3D vision / Depth and stereo
PatchRefiner V2: Fast and Lightweight Real-Domain High-Resolution Metric Depth Estimation
Zhenyu Li, Wenqing Cui, Shariq Farooq Bhat, Peter Wonka
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 end-to-end training (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.214 -> 1.184 | Improved |
| add C2F module (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.184 -> 1.041 | Improved |
| add noisy pretraining (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.041 -> 1.003 | Improved |
| scale up F2C parameters (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.201 -> 1.214 | No clear effect |
| add GDU in C2F (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.137 -> 1.003 | Improved |
| use GDU instead of PatchRefiner fusion (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.202 -> 1.003 | Improved |
| pretrain C2F and F2C in addition to the encoder (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.029 -> 1.003 | Improved |
| add ImageNet pretraining of the encoder (ablation) | PRV2-M · UnrealStereo4K | RMSE: 1.059 -> 1.003 | Improved |
| replace SSI loss with SSIGM | PR · CityScapes | boundary F1: 30.02 -> 35.19 | Improved |
| replace SSI loss with SSIGM | PRV2-E · CityScapes | boundary F1: 29.22 -> 35.32 | Improved |
| add SSIGM loss | PR · CityScapes | scale delta1: 93.308 -> 93.277 | No clear effect |
| apply SSI alignment before gradient matching (SSIGM) rather than after (GMSSI) (ablation) | PRV1 · CityScapes | EdgeComp: 14.89 -> 14.70 | Improved |
Closest papers by meaning
- PatchFusion: An End-to-End Tile-Based Framework for High-Resolution Monocular Metric Depth Estimation (CVPR 2024 · similarity 0.89)
- SharpDepth: Sharpening Metric Depth Predictions Using Diffusion Distillation (CVPR 2025 · similarity 0.86)
- Self-Distilled Depth Refinement with Noisy Poisson Fusion (NeurIPS 2024 · similarity 0.85)
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.85)
- Depth Anything V2 (NeurIPS 2024 · similarity 0.84)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.84)
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.
