Research map / 3D vision / Human pose estimation
ManiSkill2: A Unified Benchmark for Generalizable Manipulation Skills
Jiayuan Gu, Fanbo Xiang, Xuanlin Li, Zhan Ling, Xiqiang Liu, Tongzhou Mu, Yihe Tang, Stone Tao, Xinyue Wei, Yunchao Yao, Xiaodi Yuan, Pengwei Xie and 3 more
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 |
|---|---|---|---|
| enable asynchronous rendering and the render server (ablation) | ManiSkill2 visual sample collection · PickCube, random actions | FPS: 942±19 -> 2487±24 | Improved |
| enable asynchronous rendering and the render server (ablation) | ManiSkill2 visual sample collection · PickCube, Nature CNN | FPS: 931±4 -> 2532±63 | Improved |
| use point-cloud observations instead of RGBD | DAPG+PPO · PickSingleYCB | success rate: 0.18 ± 0.07 -> 0.51 ± 0.05 | Improved |
| use point-cloud observations instead of RGBD | DAPG+PPO · PickCube | success rate: 0.91 ± 0.05 -> 0.94 ± 0.03 | No clear effect |
| replace the delta end-effector pose controller with delta joint position (ablation) | Point-cloud DAPG+PPO on PickSingleYCB · PickSingleYCB | success rate: 0.51 ± 0.05 -> 0.22±0.18 | Hurt |
| train with behavior cloning instead of DAPG+PPO | Point-cloud policy · PickCube | success rate: 0.22 ± 0.06 -> 0.94 ± 0.03 | Hurt |
| use DAPG+PPO rather than behavior cloning | Point-cloud policy · PickCube | success rate: 0.22 ± 0.06 -> 0.94 ± 0.03 | Improved |
| transfer the simulation-trained policy to the real robot | Point-cloud visual RL policy · PickCube | success rate: 91.0% -> 60.0% | Hurt |
Closest papers by meaning
- ManiSkill-HAB: A Benchmark for Low-Level Manipulation in Home Rearrangement Tasks (ICLR 2025 · similarity 0.90)
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation (ICML 2025 · similarity 0.90)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.89)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 2025 · similarity 0.89)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- VideoVLA: Video Generators Can Be Generalizable Robot Manipulators (NeurIPS 2025 · similarity 0.89)
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.
