Research map / 3D vision / Depth and stereo
Enhancing Generalization of Depth Estimation Foundation Model via Weakly-Supervised Adaptation with Regularization
Yan Huang, Yongyi Su, Xin Lin, Le Zhang, Xun Xu
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 self-training loss (ablation) | WeSTAR · Sintel-C | delta1: 60.3 -> 63.3 | Improved |
| add self-training loss (ablation) | WeSTAR · Sintel (clean) | delta1: 74.8 -> 74.9 | No clear effect |
| add weak supervision (ablation) | WeSTAR · Sintel (clean) | delta1: 74.8 -> 77.5 | Improved |
| add weak supervision (ablation) | WeSTAR · Sintel (clean) | AbsRel: 20.3 -> 24.1 | Hurt |
| add weight regularization to self-training plus weak supervision (ablation) | WeSTAR · Sintel-C | delta1: 70.1 -> 71.8 | Improved |
| add weight regularization to self-training plus weak supervision (ablation) | WeSTAR · Sintel (clean) | AbsRel: 21.3 -> 16.9 | Improved |
| use LoRA instead of full-parameter tuning (ablation) | WeSTAR · Sintel (clean) | delta1: 80.8 -> 82.2 | Improved |
| use LoRA instead of full-parameter tuning (ablation) | WeSTAR · Sintel-C | delta1: 72.4 -> 71.8 | Hurt |
| tune decoder only instead of LoRA (ablation) | WeSTAR · Sintel-C | delta1: 71.8 -> 60.3 | Hurt |
| tune encoder and decoder instead of encoder only (ablation) | WeSTAR · Sintel | delta1: 80.7 -> 80.8 | No clear effect |
| use SA-HDN instead of global normalization (ablation) | WeSTAR · Sintel-C | delta1: 70.9 -> 71.8 | Improved |
| use SA-HDN instead of grid-based HDN (ablation) | WeSTAR · Sintel (clean) | delta1: 78.6 -> 82.2 | Improved |
Closest papers by meaning
- Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data (CVPR 2024 · similarity 0.88)
- Hierarchical Normalization for Robust Monocular Depth Estimation (NeurIPS 2022 · similarity 0.88)
- DME: Unveiling the Bias for Better Generalized Monocular Depth Estimation (AAAI 2024 · similarity 0.87)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.87)
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.87)
- Mind The Edge: Refining Depth Edges in Sparsely-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.87)
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.
