Research map / 3D vision / Depth and stereo
DDRNet: Depth Map Denoising and Refinement for Consumer Depth Cameras Using Cascaded CNNs
Shi Yan, Chenglei Wu, Lizhen Wang, Feng Xu, Liang An, Kaiwen Guo, Yebin Liu
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 the denoising net in front of the refinement net (cascade) instead of a bilateral filter (ablation) | DDRNet · K1 near-GT set | MAE (mm): 15.7 -> 10.9 | Improved |
| add the denoising net in front of the refinement net (cascade) instead of a bilateral filter (ablation) | DDRNet · K1 GT set | MAE (mm): 15.8 -> 11.0 | Improved |
| supervise fidelity loss with the reference depth rather than the denoised depth only (ablation) | DDRNet · K1 near-GT set | MAE (mm): 16.1 -> 10.9 | Improved |
| supervise fidelity loss with the reference depth rather than the denoised depth only (ablation) | DDRNet · K1 GT set | RMSE (mm): 5.5 -> 3.6 | Improved |
| add the refinement net after the denoising net (Ddt vs Ddn output) (ablation) | DDRNet · K2 validation, normal difference | percent within 10.0 degrees: 84.69 -> 83.69 | No clear effect |
| add the refinement net after the denoising net (Ddt vs Ddn output) (ablation) | DDRNet · K2 validation, depth difference seq. 1 | depth difference: 19.03 -> 18.97 | No clear effect |
| add the denoising CNN to the pipeline (ablation) | DDRNet · runtime (TitanX, 640x480) | ms per frame: 9.6 -> 20.4 | Hurt |
Later papers that compared with this method
5 later accepted papers report DDRNet as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Attention-Based Multi-Modal Fusion Network for Semantic Scene Completion (AAAI 2020 · better on 1 of 2 shared results)
- NormalGAN: Learning Detailed 3D Human from a Single RGB-D Image (ECCV 2020 · better on 2 of 2 shared results)
- Anisotropic Convolutional Networks for 3D Semantic Scene Completion (CVPR 2020 · better on 1 of 1 shared result)
- Cascaded Context Pyramid for Full-Resolution 3D Semantic Scene Completion (ICCV 2019 · better on 4 of 4 shared results)
- Self-Supervised Deep Depth Denoising (ICCV 2019 · 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
- Self-Supervised Deep Depth Denoising (ICCV 2019 · similarity 0.87)
- Spatial Hierarchy Aware Residual Pyramid Network for Time-of-Flight Depth Denoising (ECCV 2020 · similarity 0.87)
- Self-Distilled Depth Refinement with Noisy Poisson Fusion (NeurIPS 2024 · similarity 0.86)
- Refine and Distill: Exploiting Cycle-Inconsistency and Knowledge Distillation for Unsupervised Monocular Depth Estimation (CVPR 2019 · similarity 0.85)
- DepthInSpace: Exploitation and Fusion of Multiple Video Frames for Structured-Light Depth Estimation (ICCV 2021 · similarity 0.85)
- DORNet: A Degradation Oriented and Regularized Network for Blind Depth Super-Resolution (CVPR 2025 · similarity 0.85)
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.
