Research map / 3D vision / Depth and stereo
ProDepth: Boosting Self-Supervised Multi-Frame Monocular Depth with Probabilistic Fusion
Sungmin Woo, Wonjoon Lee, Woo Jin Kim, Dogyoon Lee, Sangyoun Lee
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 probabilistic cost volume modulation (PCVM) with weighted uncertainty (ablation) | ProDepth · Cityscapes | Abs Rel: 0.100 -> 0.095 | Improved |
| add PCVM on top of segmentation masks (ablation) | ProDepth · Cityscapes | Abs Rel: 0.100 -> 0.101 | Hurt |
| add PCVM on top of ManyDepth consistency mask (ablation) | ProDepth · Cityscapes | Abs Rel: 0.107 -> 0.103 | Improved |
| use weighted (probabilistic) uncertainty instead of binary uncertainty with PCVM (ablation) | ProDepth · Cityscapes | Abs Rel: 0.099 -> 0.098 | Improved |
| use loss reweighting instead of binary masking (ablation) | ProDepth · Cityscapes | Abs Rel: 0.098 -> 0.097 | Improved |
| combine masking and reweighting (ablation) | ProDepth · Cityscapes | Abs Rel: 0.097 -> 0.095 | Improved |
| use weighted geometric mean instead of weighted arithmetic mean for fusing distributions (ablation) | PCVM · Cityscapes | Abs Rel: 0.098 -> 0.095 | Improved |
| use learned weighted uncertainty instead of segmentation mask (no PCVM) (ablation) | ProDepth · Cityscapes | Abs Rel: 0.101 -> 0.100 | No clear effect |
| use ProDepth without semantics vs DynamicDepth with semantics on Cityscapes | multi-frame depth · Cityscapes | Abs Rel: 0.103 -> 0.095 | Improved |
| apply ProDepth to Waymo Open after Cityscapes pretraining | multi-frame depth · Waymo Open | Abs Rel: 0.255 -> 0.247 (vs DynamicDepth) | Improved |
Later papers that compared with this method
1 later accepted paper reports ProDepth as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Towards Robust Driving Perception: A Flexible Scale-Driven Family for Self-Supervised Monocular Depth Estimation (ECCV 2026 · 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
- AdaDepth: Exploiting Inherent Scene Information for Self-Supervised Depth Estimation in Dynamic Scenes (AAAI 2026 · similarity 0.92)
- The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth (CVPR 2021 · similarity 0.92)
- Mining Supervision for Dynamic Regions in Self-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.90)
- SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation (AAAI 2024 · similarity 0.90)
- GeoDepth: From Point-to-Depth to Plane-to-Depth Modeling for Self-Supervised Monocular Depth Estimation (CVPR 2025 · similarity 0.89)
- From-Ground-To-Objects: Coarse-to-Fine Self-supervised Monocular Depth Estimation of Dynamic Objects with Ground Contact Prior (CVPR 2024 · similarity 0.89)
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.
