Research map / 3D vision / Depth and stereo
CLIFFNet for Monocular Depth Estimation with Hierarchical Embedding Loss
Lijun Wang, Jianming Zhang, Yifan Wang, Huchuan Lu, Xiang Ruan
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 HEG-S hierarchical embedding loss to L1 loss (ablation) | CLIFFNet · NYU-Depth V2 (CLIFFNet) | RMSE: 0.529 -> 0.493 | Improved |
| add HEG-R hierarchical embedding loss to L1 loss (ablation) | CLIFFNet · NYU-Depth V2 (CLIFFNet) | RMSE: 0.529 -> 0.497 | Improved |
| add HEG-S hierarchical embedding loss to L1 loss (ablation) | CLIFFNet · NYU-Depth V2 (CLIFFNet) | P1: 0.817 -> 0.844 | Improved |
| use HEG-S loss instead of depth gradient loss (ablation) | CLIFFNet · NYU-Depth V2 (CLIFFNet) | RMSE: 0.513 -> 0.493 | Improved |
| use HEG-S loss instead of depth aware loss (ablation) | CLIFFNet · NYU-Depth V2 (CLIFFNet) | RMSE: 0.511 -> 0.493 | Improved |
| train HEG on relevant task rather than randomly initialize it (ablation) | HEG-R loss · NYU-Depth V2 (CLIFFNet) | RMSE: 0.517 -> 0.497 | Improved |
| train HEG on depth-relevant task rather than ImageNet classification (ablation) | hierarchical loss · NYU-Depth V2 (CLIFFNet) | RMSE: 0.523 -> 0.493 | Improved |
| add scene classification as multi-task head instead of HEG-S loss (ablation) | CLIFFNet with L1 · NYU-Depth V2 (CLIFFNet) | RMSE: 0.529 -> 0.530 | No clear effect |
| use scale invariant loss in addition to L1 (ablation) | CLIFFNet · NYU-Depth V2 (CLIFFNet) | RMSE: 0.529 -> 0.520 | Improved |
| add attention-based low-level feature refinement to CLIFF module (ablation) | CLIFFNet with HEG-S · NYU-Depth V2 (CLIFFNet) | RMSE, Abs Rel, P1, P2 | Improved |
| add multi-level feature selection to CLIFF module (ablation) | CLIFFNet with HEG-S · NYU-Depth V2 (CLIFFNet) | RMSE, Abs Rel, P1, P2 | Improved |
| add identity mapping of high-level features to CLIFF module (ablation) | CLIFFNet with HEG-S · NYU-Depth V2 (CLIFFNet) | RMSE, Abs Rel, P1, P2 | Improved |
Closest papers by meaning
- BRNet: Exploring Comprehensive Features for Monocular Depth Estimation (ECCV 2022 · similarity 0.90)
- Deep Digging into the Generalization of Self-Supervised Monocular Depth Estimation (AAAI 2023 · similarity 0.88)
- DeFeat-Net: General Monocular Depth via Simultaneous Unsupervised Representation Learning (CVPR 2020 · similarity 0.88)
- Depth from a Single Image by Harmonizing Overcomplete Local Network Predictions (NeurIPS 2016 · similarity 0.88)
- MonoMVSNet: Monocular Priors Guided Multi-View Stereo Network (ICCV 2025 · similarity 0.88)
- HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation (AAAI 2021 · 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.
