Research map / Reinforcement learning / Visual and representation RL
SAPIEN: A SimulAted Part-Based Interactive ENvironment
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use RGB-D instead of RGB for Mask R-CNN (ablation) | Mask R-CNN · PartNet-Mobility test set | mAP (AP%): 53.0 -> 52.8 | No clear effect |
| Add RGB features to XYZ point-cloud input (ablation) | PartNet InsSeg · PartNet-Mobility test set | mAP (AP%): 36.1 -> 37.1 | Helped |
| Use ResNet-50 RGB-D input instead of PointNet++ RGB point cloud (ablation) | Motion attribute estimation · Doors and drawers | drawer pose error: 0.0918 -> 0.0645 | Helped |
| Use mobility-based state representation instead of raw scene state (ablation) | SAC door opening · Unseen doors, train 16 objects | final angle (degrees): 27.3 -> 32.8 | Helped |
| Use visual state representation instead of raw scene state (ablation) | SAC door opening · Unseen doors, train 16 objects | final angle (degrees): 27.3 -> 29.6 | Helped |
| Use mobility-based state representation instead of visual input (ablation) | SAC drawer opening · Unseen drawers, train 8 objects | success rate: 0.61 -> 0.69 | Helped |
| Train on more object instances (ablation) | SAC door opening · Unseen doors | final angle (degrees): 14.7 -> 27.3 | Helped |
| Train on more object instances (ablation) | SAC door opening · Training doors | final angle (degrees): 85.4 -> 38.4 | Hurt |
| Train on more object instances (ablation) | SAC drawer opening · Unseen drawers | success rate: 0.61 -> 0.66 | Helped |
| Use visual inputs rather than mobility-based state (ablation) | SAC drawer opening · Unseen drawers | success rate: 0.69 -> 0.61 | Hurt |
| Use a heuristic controller with ground-truth segmentation | Drawer-pulling · 108 cabinet instances | success rate: 95.3% | Helped |
| Use PBVS to track and clamp the door edge after initial opening | Door-opening · 77 cabinet instances | success rate: 81.8% | Helped |
Closest papers by meaning
- AdaManip: Adaptive Articulated Object Manipulation Environments and Policy Learning (ICLR 2025 · similarity 0.84)
- Learning Interactive Real-World Simulators (ICLR 2024 · similarity 0.83)
- Habitat 3.0: A Co-Habitat for Humans, Avatars, and Robots (ICLR 2024 · similarity 0.83)
- ManiTaskGen: A Comprehensive Task Generator for Benchmarking and Improving Vision-Language Agents on Embodied Decision-Making (NeurIPS 2026 · similarity 0.83)
- RobotSmith: Generative Robotic Tool Design for Acquisition of Complex Manipulation Skills (NeurIPS 2025 · similarity 0.83)
- PARTNR: A Benchmark for Planning and Reasoning in Embodied Multi-agent Tasks (ICLR 2025 · similarity 0.83)
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.