Research map / 3D vision / Human pose estimation
Breaking the 3D Dataset Bottleneck: Fast Scalable Generation of Aligned 3D Assets from Scratch for Category 6D Pose Estimation and Robotic Grasping
Duret Guillaume, Danylo Mazurak, Florence Zara, Jan Peters, Liming Chen
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 depth conditioning (ControlNet) to text-to-image generation (ablation) | text-to-image-to-3D pipeline · NOCS categories | pose consistency (%): 57 -> 97 | Improved |
| add depth conditioning for laptops (ablation) | image generation · NOCS Laptop | pose consistency (%): 20 -> 90 | Improved |
| add shadows to mixed-reality training data (generated meshes) (ablation) | DualPoseNet · NOCS REAL275 zero-shot | average pose score: 30.83 -> 34.75 | Improved |
| add shadows to mixed-reality training data (original NOCS meshes) (ablation) | DualPoseNet · NOCS REAL275 zero-shot | average pose score: 29.07 -> 33.24 | Improved |
| replace original NOCS meshes with generated meshes (with shadows) (ablation) | DualPoseNet · NOCS REAL275 zero-shot | average pose score: 33.24 -> 34.75 | Improved |
| replace original NOCS meshes with generated meshes (no shadows) (ablation) | DualPoseNet · NOCS REAL275 zero-shot | average pose score: 29.07 -> 30.83 | Improved |
| replace original NOCS meshes with generated meshes (ablation) | DualPoseNet · synthetic validation | average pose score: 15.66 -> 23.91 | Improved |
| use mixed-reality rendering instead of fully synthetic Replica scenes (ablation) | DualPoseNet · NOCS REAL275 zero-shot | average pose score: 33.10 -> 34.75 | Improved |
| train CenterGrasp on generated custom data instead of pretrained CenterGrasp | CenterGrasp · SAPIEN grasping (textured eval) | grasp success rate: 0.7896 -> 0.8679 | Improved |
| train on generated custom data | CenterGrasp · SAPIEN shape completion (textured eval) | IoU: 0.314 -> 0.475 | Improved |
| train with native textures instead of randomized textures, tested on randomized textures (ablation) | Custom-CG · SAPIEN grasping (randomized eval) | grasp success rate: 0.8784 -> 0.8264 | Hurt |
| use specialized category-fine-tuned model instead of baseline CenterGrasp | CenterGrasp · real robot, Bowl | grasp success (%): 14% -> 57% | Improved |
Closest papers by meaning
- CAP-Net: A Unified Network for 6D Pose and Size Estimation of Categorical Articulated Parts from a Single RGB-D Image (CVPR 2025 · similarity 0.89)
- Pose6DAug: Physically Plausible Multi-View Object Swapping for Robot Data Augmentation (NeurIPS 2026 · similarity 0.88)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · similarity 0.88)
- RecDreamer: Consistent Text-to-3D Generation via Uniform Score Distillation (ICLR 2025 · similarity 0.87)
- ZeroGrasp: Zero-Shot Shape Reconstruction Enabled Robotic Grasping (CVPR 2025 · similarity 0.87)
- UA-Pose: Uncertainty-Aware 6D Object Pose Estimation and Online Object Completion with Partial References (CVPR 2025 · 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.
