Research map / Point clouds / Shape and surface completion
3D Implicit Transporter for Temporally Consistent Keypoint Discovery
Chengliang Zhong, Yuhang Zheng, Yupeng Zheng, Hao Zhao, Li Yi, Xiaodong Mu, Ling Wang, Pengfei Li, Guyue Zhou, Chao Yang, Xinliang Zhang, Jian Zhao
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 |
|---|---|---|---|
| replace point-based TopNet decoder with implicit reconstruction (ablation) | 3D Implicit Transporter · PartNet-Mobility test categories | relative repeatability: 0.170 -> 0.457 | Improved |
| add cross-attention for feature fusion (ablation) | 3D Implicit Transporter · PartNet-Mobility test categories | relative repeatability: 0.457 -> 0.530 | Improved |
| add correspondence loss L_corr (ablation) | 3D Implicit Transporter · PartNet-Mobility test categories | relative repeatability: 0.530 -> 0.602 | Improved |
| add axis loss L_axis (ablation) | 3D Implicit Transporter · PartNet-Mobility test categories | ADD: 0.123 -> 0.109 | Improved |
| increase number of random negative queries from 1k to 5k (ablation) | 3D Implicit Transporter · PartNet-Mobility | repeatability: 0.597 -> 0.663 | Improved |
| sample negative queries uniformly rather than randomly (ablation) | 3D Implicit Transporter · PartNet-Mobility | repeatability: 0.611 -> 0.058 | Hurt |
| test without a mask to filter body keypoints | baselines versus Ours · ITOP | Chamfer distance: ISS 0.29 -> 0.89; Ours 0.13 -> 0.14 | Hurt |
| use keypoint-based manipulation versus UMPNet on test categories | goal-conditioned manipulation · PartNet-Mobility test categories | success rate: 0.79 -> 0.83 | Improved |
Closest papers by meaning
- KeyTr: Keypoint Transporter for 3D Reconstruction of Deformable Objects in Videos (CVPR 2022 · similarity 0.88)
- Unsupervised Learning of Object Keypoints for Perception and Control (NeurIPS 2019 · similarity 0.88)
- KeypointDeformer: Unsupervised 3D Keypoint Discovery for Shape Control (CVPR 2021 · similarity 0.88)
- Unsupervised Monocular 3D Keypoint Discovery from Multi-View Diffusion Priors (CVPR 2026 · similarity 0.87)
- Key-Grid: Unsupervised 3D Keypoints Detection using Grid Heatmap Features (NeurIPS 2024 · similarity 0.85)
- PointZero: 3D Point Track Completion for Learning Transferable 3D Dynamics (NeurIPS 2026 · 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.
