AI Paper Atlas

Research map / 3D vision / 3D generation

Unsupervised Learning of Visual 3D Keypoints for Control

Boyuan Chen, Pieter Abbeel, Deepak Pathak

ICML 2021 · 3D vision › 3D generation · approach: 2D-to-3D pretrained representations · 11 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
Use 3D keypoints instead of RAD's representationVisual RL on Meta-World · Meta-World (8 environments)RL performanceImproved
Use 3D keypoints instead of visual RL baselinesVisual RL for locomotion · PyBullet Ant and colorless AntRL performanceImproved
Use 3D keypoints instead of prior visual RL methodsScarf manipulation · Scarf manipulationRL performanceImproved
Use 3D keypoints rather than baseline representationsTransfer learning · ML45 training environmentsRL performanceImproved
Use a learned representation instead of RAD's representationTransfer learning · ML45 test environmentstransfer performanceImproved
Remove random crop offset (ablation)Keypoint3D · Meta-World open doorRL performanceHurt
Remove multi-view consistency loss (ablation)Keypoint3D · Meta-World ablation environmentsRL performanceHurt
Use triangulation to map 2D keypoints into 3D instead of the learned 3D keypoint method (ablation)Keypoint3D · Meta-World ablation environmentsRL performanceHurt
Train keypoints on random-policy data, then freeze the encoder for policy learning (ablation)Keypoint3D · Meta-World joint-training ablationRL performanceHurt
Train keypoints on expert-policy data, then freeze the encoder for policy learning (ablation)Keypoint3D · Meta-World joint-training ablationRL performanceHurt
Stop policy gradients from updating the encoder (ablation)Keypoint3D · Meta-World joint-training ablationRL performanceHurt
Remove attention (ablation)Keypoint3D · Reinforcement learning ablation environmentsRL performanceHurt

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