Research map / 3D vision / Depth and stereo
Domain Randomization-Enhanced Depth Simulation and Restoration for Perceiving and Grasping Specular and Transparent Objects
Qiyu Dai, Jiyao Zhang, Qiwei Li, Tianhao Wu, Hao Dong, Ziyuan Liu, Ping Tan, He Wang
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 |
|---|---|---|---|
| train on simulated DREDS depth instead of cropped synthetic depth (ablation) | SwinDRNet · STD-CatKnown | RMSE: 0.0467 -> 0.015 | Improved |
| train on simulated DREDS depth instead of cropped synthetic depth (ablation) | SwinDRNet · STD-CatNovel | RMSE: 0.065 -> 0.025 | Improved |
| train on DREDS-CatKnown and transfer to ClearGrasp instead of training on ClearGrasp and Omniverse | SwinDRNet · ClearGrasp real-known | RMSE: 0.028 -> 0.022 | Improved |
| train on DREDS-CatKnown and transfer to ClearGrasp instead of training on ClearGrasp and Omniverse | SwinDRNet · ClearGrasp real-novel | RMSE: 0.025 -> 0.016 | Improved |
| add RGB-D fusion module (ablation) | SwinDRNet · STD-CatKnown | RMSE: 0.019 -> 0.014 | Improved |
| add confidence interpolation module (ablation) | SwinDRNet · STD-CatKnown | RMSE: 0.014 -> 0.015 | No clear effect |
| add confidence interpolation module (ablation) | SwinDRNet+NOCSHead · STD-CatKnown | 5deg 5cm pose accuracy: 52.2 -> 54.1 | Improved |
| add material randomization instead of fixed ShapeNet materials (ablation) | SwinDRNet · STD-CatKnown | RMSE: 0.024 -> 0.015 | Improved |
| add material randomization instead of fixed ShapeNet materials (ablation) | SwinDRNet · STD-CatNovel | RMSE: 0.038 -> 0.025 | Improved |
| double training data scale from half to full DREDS (ablation) | SwinDRNet · STD-CatKnown | RMSE: 0.021 -> 0.018 | Improved |
| double training data scale from half to full DREDS (ablation) | SwinDRNet · STD-CatNovel | RMSE: 0.028 -> 0.025 | Improved |
| use restored depth from SwinDRNet for pose fitting instead of raw depth (ablation) | SwinDRNet+NOCSHead · STD-CatKnown | 5deg 5cm pose accuracy: 42.9 -> 54.1 | Improved |
Closest papers by meaning
- Consistent Depth Prediction for Transparent Object Reconstruction from RGB-D Camera (ICCV 2023 · similarity 0.88)
- RGB-D Local Implicit Function for Depth Completion of Transparent Objects (CVPR 2021 · similarity 0.87)
- Deep Depth Completion of a Single RGB-D Image (CVPR 2018 · similarity 0.87)
- RestorGS: Depth-aware Gaussian Splatting for Efficient 3D Scene Restoration (CVPR 2025 · similarity 0.86)
- MVD-Fusion: Single-view 3D via Depth-consistent Multi-view Generation (CVPR 2024 · similarity 0.85)
- Robust Depth Enhancement via Polarization Prompt Fusion Tuning (CVPR 2024 · 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.
