Research map / 3D vision / Human pose estimation
Local Occupancy-Enhanced Object Grasping with Multiple Triplanar Projection
Kangqi Ma, Hao Dong, Yadong Mu
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 local occupancy prediction with multi-group tri-plane to GSNet-style grasp estimation | grasp pose estimator · GraspNet-1Billion Similar (RealSense) | AP: 53.75 -> 63.41 | Improved |
| filter grasp poses colliding with predicted occupancy (collision detection) (ablation) | Ours · GraspNet-1Billion Avg (RealSense) | AP: 53.87 -> 55.56 | Improved |
| add occupancy enhancement versus none (ablation) | grasp pose estimator · GraspNet-1Billion Avg (RealSense) | AP: 49.72 -> 53.87 | Improved |
| predict occupancy locally instead of whole scene with multi-group tri-plane (ablation) | Ours · GraspNet-1Billion Avg (RealSense) | AP and time: AP 51.75 -> 53.87, time 0.35s -> 0.18s | Improved |
| use ball query local context only without tri-plane global context (ablation) | Ours · GraspNet-1Billion Avg (RealSense) | AP: 53.87 -> 50.60 | Hurt |
| add tri-plane point embeddings to grasp model (ablation) | Ours · GraspNet-1Billion Avg (RealSense) | AP: 49.72 -> 52.03 | Improved |
| add point density to tri-plane (ablation) | Ours · GraspNet-1Billion Avg (RealSense) | AP: 52.03 -> 52.16 | No clear effect |
| add local context to occupancy query (ablation) | Ours · GraspNet-1Billion Avg (RealSense) | AP: 52.16 -> 53.39 | Improved |
| add grasp direction refinement from completed shape (ablation) | Ours · GraspNet-1Billion Avg (RealSense) | AP: 53.39 -> 53.87 | Improved |
| increase number of tri-plane groups K from 1 to 3 (ablation) | Ours · GraspNet-1Billion Similar | AP0.2: 19.85 -> 23.77 | Improved |
| increase tri-plane resolution H and W | Ours · GraspNet-1Billion | AP | No clear effect |
| deploy on a real robot arm versus GSNet baseline | Ours · real-world cluttered grasping | success rate: +6.62% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GSNet — GSNet: Joint Vehicle Pose and Shape Reconstruction with Geometrical and Scene-aware Supervision (ECCV 2020) · better on 5 of 5 shared results
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
- ZeroGrasp: Zero-Shot Shape Reconstruction Enabled Robotic Grasping (CVPR 2025 · similarity 0.88)
- BOP-ASK: Object-Interaction Reasoning for Vision-Language Models (CVPR 2026 · similarity 0.87)
- CADGrasp: Learning Contact and Collision Aware General Dexterous Grasping in Cluttered Scenes (NeurIPS 2025 · similarity 0.86)
- Single-View Scene Point Cloud Human Grasp Generation (CVPR 2024 · similarity 0.86)
- Towards Flexible 3D Perception: Object-Centric Occupancy Completion Augments 3D Object Detection (NeurIPS 2024 · similarity 0.86)
- GEARS: Local Geometry-aware Hand-object Interaction Synthesis (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.
