Research map / 3D vision / Human pose estimation
RAVEN: End-to-end Equivariant Robot Learning with RGB Cameras
David Klee, Boce Hu, Andrew Cole, Heng Tian, Dian Wang, Robert Platt, Robin Walters
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 SE(3) ray encoding (patch orientation) (ablation) | RAVEN · MimicGen ablation subset (4 tasks) | success rate (%): 72 -> 76 | Improved |
| add equivariant decoder (ablation) | RAVEN · MimicGen ablation subset (4 tasks) | success rate (%): 72 -> 76 | Improved |
| add equivariant encoder and decoder (ablation) | RAVEN · MimicGen ablation subset (4 tasks) | success rate (%): 58 -> 76 | Improved |
| use RAVEN instead of EquiDiffPo | policy · MimicGen (12 tasks, 100 demos) | success rate (%): 54 -> 66 | Improved |
| use RAVEN architecture over DiffPo with same pretrained ResNet | DiffPo (Pre) · MimicGen (12 tasks, 100 demos) | success rate (%): 52 -> 66 | Improved |
| use pretrained ResNet encoder in DiffPo | DiffPo · MimicGen (12 tasks, 100 demos) | success rate (%): 42 -> 52 | Improved |
| use RAVEN on bimanual tasks | policy · DexMimicGen (6 tasks, 50 demos) | success rate (%): 65 -> 82 | Improved |
| use RAVEN under perturbed camera views | policy · MimicGen perturbed agent-view camera (4 tasks) | success rate (%): 48 -> 71 | Improved |
| use RAVEN on real-world tasks | policy · Real-world UR5 (4 tasks) | success rate (%): 24 -> 63 | Improved |
| use RAVEN instead of EquiDiffPo for training speed | policy · MimicGen Threading D2 | training time (hours): 7.3 -> 2.8 | Improved |
| add camera-parameter augmentation at test time under calibration noise | RAVEN · MimicGen subset (calibration noise) | success rate (%): dropping only 1% | Improved |
Closest papers by meaning
- 3D Equivariant Visuomotor Policy Learning via Spherical Projection (NeurIPS 2025 · similarity 0.90)
- ET-SEED: EFFICIENT TRAJECTORY-LEVEL SE(3) EQUIVARIANT DIFFUSION POLICY (ICLR 2025 · similarity 0.88)
- Hierarchical Equivariant Policy via Frame Transfer (ICML 2025 · similarity 0.88)
- Deep SE(3)-Equivariant Geometric Reasoning for Precise Placement Tasks (ICLR 2024 · similarity 0.87)
- Learning to Act from Actionless Videos through Dense Correspondences (ICLR 2024 · similarity 0.86)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.86)
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.
