Research map / 3D vision / Depth and stereo
Consistent Depth Prediction for Transparent Object Reconstruction from RGB-D Camera
Yuxiang Cai, Yifan Zhu, Haiwei Zhang, Bo Ren
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 surface normal prediction branch during training (ablation) | stereo transparent depth network · Synthetic transparent-object video dataset (validation) | RMSE: 0.178 -> 0.148 | Improved |
| use ground-truth camera trajectories in training instead of estimated ones (ablation) | stereo transparent depth network · Synthetic transparent-object video dataset (validation) | RMSE: 0.156 -> 0.148 | Improved |
| use two neighboring views (w=2) instead of one (ablation) | stereo transparent depth network · Synthetic transparent-object video dataset (validation) | RMSE/MAE/REL | Improved |
| increase neighboring views beyond 2 (ablation) | stereo transparent depth network · Synthetic transparent-object video dataset (validation) | accuracy vs inference time: about 0.08s extra per additional view | Mixed |
| use multi-view plane-sweep stereo depth instead of monocular transparent-depth prediction (TransCG) | transparent object reconstruction · Synthetic scene reconstruction | Chamfer distance: 0.046 -> 0.027 | Improved |
| use proposed depth prediction instead of ClearGrasp in the pseudo-SLAM pipeline | trajectory estimation · Trajectory evaluation, Goblet(part) scene | ATE: 0.0152 -> 0.0736 | Hurt |
| use proposed method instead of LIDF on delta 1.05 | transparent depth prediction · Synthetic transparent-object video dataset (validation) | delta 1.05: 53.64 -> 50.30 | Hurt |
Closest papers by meaning
- RGB-D Local Implicit Function for Depth Completion of Transparent Objects (CVPR 2021 · similarity 0.90)
- Neural RGB(r)D Sensing: Depth and Uncertainty From a Video Camera (CVPR 2019 · similarity 0.88)
- Transfusion: A Novel SLAM Method Focused on Transparent Objects (ICCV 2021 · similarity 0.88)
- Estimating Depth from RGB and Sparse Sensing (ECCV 2018 · similarity 0.88)
- Deep Depth Completion of a Single RGB-D Image (CVPR 2018 · similarity 0.88)
- Domain Randomization-Enhanced Depth Simulation and Restoration for Perceiving and Grasping Specular and Transparent Objects (ECCV 2022 · similarity 0.88)
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.
