Research map / 3D vision / Depth and stereo
Hybrid-grained Feature Aggregation with Coarse-to-fine Language Guidance for Self-supervised Monocular Depth Estimation
Wenyao Zhang, Hongsi Liu, Bohan Li, Jiawei He, Zekun Qi, Yunnan Wang, Shengyang Zhao, Xinqiang Yu, Wenjun Zeng, Xin Jin
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 |
|---|---|---|---|
| plug Hybrid-depth encoder into Monodepth2 | Monodepth2 · KITTI Eigen split | Abs Rel: 0.115 -> 0.093 | Improved |
| plug Hybrid-depth encoder into Manydepth | Manydepth · KITTI Eigen split | Abs Rel: 0.098 -> 0.096 | Improved |
| plug Hybrid-depth encoder into Mono-ViFI | Mono-ViFI · KITTI Eigen split | Abs Rel: 0.105 -> 0.094 | Improved |
| add coarse depth sensing stage (vs stronger backbone alone) (ablation) | Monodepth2 · KITTI Eigen split | Abs Rel: 0.105 -> 0.093 | Improved |
| add coarse depth sensing stage (vs stronger backbone alone) (ablation) | Manydepth · KITTI Eigen split | Abs Rel: 0.096 -> 0.096 | No clear effect |
| add language-guided contrastive loss on top of intramodal loss (ablation) | Monodepth2 + Hybrid-depth · KITTI Eigen split | Abs Rel: 0.098 -> 0.093 | Improved |
| add intramodal contrastive loss on top of language-guided loss (ablation) | Monodepth2 + Hybrid-depth · KITTI Eigen split | Abs Rel: 0.095 -> 0.093 | Improved |
| add depth-instruction language alignment (granularity calibrator) in coarse stage (ablation) | Monodepth2 + Hybrid-depth · KITTI Eigen split | Abs Rel: 0.098 -> 0.093 | Improved |
| add depth-instruction language alignment in fine stage (ablation) | Monodepth2 + Hybrid-depth · KITTI Eigen split | Abs Rel: 0.100 -> 0.093 | Improved |
| combine CLIP and DINO instead of using only one (ablation) | Monodepth2 + Hybrid-depth · KITTI Eigen split | Abs Rel: 0.102 -> 0.093 (CLIP only); 0.104 -> 0.093 (DINO only) | Improved |
| increase learnable depth tokens beyond 256 (ablation) | Monodepth2 + Hybrid-depth · KITTI Eigen split | Abs Rel: 0.093 -> 0.097 (512 tokens) | Hurt |
| plug Hybrid-depth into BEV detectors | BEVDet and FB-BEV · nuScenes val | mAP / NDS: BEVDet mAP 0.283 -> 0.325; FB-BEV NDS 0.406 -> 0.439 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SQLdepth — SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation (AAAI 2024) · better on 1 of 1 shared result
- FB-BEV — FB-BEV: BEV Representation from Forward-Backward View Transformations (ICCV 2023) · 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
- Fine-Grained Semantics-Aware Representation Enhancement for Self-Supervised Monocular Depth Estimation (ICCV 2021 · similarity 0.92)
- Towards Comprehensive Representation Enhancement in Semantics-Guided Self-Supervised Monocular Depth Estimation (ECCV 2022 · similarity 0.90)
- Excavating the Potential Capacity of Self-Supervised Monocular Depth Estimation (ICCV 2021 · similarity 0.90)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.90)
- Self-Supervised Monocular Depth Estimation by Direction-aware Cumulative Convolution Network (ICCV 2023 · similarity 0.89)
- SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation (AAAI 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.
