Research map / Vision-language and multimodal / Robot action and navigation
Exo2EgoPolicy: Pose-Aligned Cross-View Policy Learning
Hritam Basak, Hadi Tabatabaee, Xin Yang, Shreekant Gayaka, Nan Qiao, Yuyin Sun, Cheng-Hao Kuo, Zhaozheng Yin, Min Sun
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 |
|---|---|---|---|
| translate in pose space instead of pixel-space synthesis (vs EgoX) | Exo2EgoPolicy · H2O unseen objects (exo-to-ego pose) | PA-MPJPE (mm): 15.7 -> 11.4 | Improved |
| pose-space translation vs pixel-space translation co-training (EgoX) | Exo2EgoPolicy · Ego Humanoid Manipulation Benchmark, short-horizon (Unseen) | mean success rate (%): 66.6 -> 73.7 | Improved |
| translate exo poses to ego before co-training instead of using raw exo poses | Exo2EgoPolicy · Ego Humanoid Manipulation Benchmark, short-horizon (Seen) | mean success rate (%): 65.7 -> 81.9 | Improved |
| add raw exocentric poses without view translation to co-training | EgoVLA-NoPretrain · Ego Humanoid Manipulation Benchmark, short-horizon (Seen) | mean success rate (%): 64.6 -> 65.7 | No clear effect |
| exo-to-ego pose translation instead of egocentric video pretraining (EgoVLA) | Exo2EgoPolicy · Ego Humanoid Manipulation Benchmark, long-horizon (Seen) | mean success rate (%): 45.9 -> 53.3 | Improved |
| use object-relative coordinate frame (ablation) | Exo2EgoPolicy · Ablation: H2O unseen objects pose / Ego Humanoid Unseen policy | PA-MPJPE (mm): 18.4 -> 11.4 | Improved |
| use object-relative coordinate frame (ablation) | Exo2EgoPolicy · Ablation: H2O unseen objects pose / Ego Humanoid Unseen policy | short-horizon SR (%): 58.3 -> 73.7 | Improved |
| use geometric latent encoding instead of direct regression (ablation) | Exo2EgoPolicy · Ablation: H2O unseen objects pose / Ego Humanoid Unseen policy | PA-MPJPE (mm): 21.6 -> 11.4 | Improved |
| condition score network on previous pose p_{t-1} (ablation) | Exo2EgoPolicy · Ablation: H2O unseen objects pose / Ego Humanoid Unseen policy | short-horizon SR (%): 62.4 -> 73.7 | Improved |
| add temporal consistency loss L_temp (ablation) | Exo2EgoPolicy · Ablation: H2O unseen objects pose / Ego Humanoid Unseen policy | long-horizon SR (%): 19.8 -> 36.1 | Improved |
Closest papers by meaning
- OASIS: Observation-Action Space Alignment via SE(3) Trajectory Prediction for Robotic Manipulation (NeurIPS 2026 · similarity 0.89)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.88)
- DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation (NeurIPS 2026 · similarity 0.88)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.88)
- SAM-E: Leveraging Visual Foundation Model with Sequence Imitation for Embodied Manipulation (ICML 2024 · similarity 0.88)
- Pose6DAug: Physically Plausible Multi-View Object Swapping for Robot Data Augmentation (NeurIPS 2026 · similarity 0.88)
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.
