AI Paper Atlas

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

ICLR 2023 · 3D vision › Human pose estimation · approach: Category-level pose estimation · 22 citations (OpenAlex)

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 changedWhereMeasuredEffect
enable asynchronous rendering and the render server (ablation)ManiSkill2 visual sample collection · PickCube, random actionsFPS: 942±19 -> 2487±24Improved
enable asynchronous rendering and the render server (ablation)ManiSkill2 visual sample collection · PickCube, Nature CNNFPS: 931±4 -> 2532±63Improved
use point-cloud observations instead of RGBDDAPG+PPO · PickSingleYCBsuccess rate: 0.18 ± 0.07 -> 0.51 ± 0.05Improved
use point-cloud observations instead of RGBDDAPG+PPO · PickCubesuccess rate: 0.91 ± 0.05 -> 0.94 ± 0.03No clear effect
replace the delta end-effector pose controller with delta joint position (ablation)Point-cloud DAPG+PPO on PickSingleYCB · PickSingleYCBsuccess rate: 0.51 ± 0.05 -> 0.22±0.18Hurt
train with behavior cloning instead of DAPG+PPOPoint-cloud policy · PickCubesuccess rate: 0.22 ± 0.06 -> 0.94 ± 0.03Hurt
use DAPG+PPO rather than behavior cloningPoint-cloud policy · PickCubesuccess rate: 0.22 ± 0.06 -> 0.94 ± 0.03Improved
transfer the simulation-trained policy to the real robotPoint-cloud visual RL policy · PickCubesuccess rate: 91.0% -> 60.0%Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy