Research map / 3D vision / Human pose estimation
Robot Structure Prior Guided Temporal Attention for Camera-to-Robot Pose Estimation From Image Sequence
Yang Tian, Jiyao Zhang, Zekai Yin, Hao Dong
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 |
|---|---|---|---|
| Use successive frames with robot structure-guided feature alignment and temporal attention | camera-to-robot pose estimation · Panda Orb | ADD AUC: 52.56 -> 60.30 | Improved |
| Add structure-guided feature alignment and reprojection belief map (ablation) | SGTAPose · Panda Orb | PCK AUC: 49.78 -> 62.86 | Improved |
| Add temporal cross-attention to aligned frame features (ablation) | SGTAPose with structure-guided alignment · Panda Orb | ADD AUC: 55.83 -> 58.86 | Improved |
| Add the pose refiner with reweighted PnP optimization (ablation) | SGTAPose with structure-guided alignment and temporal cross-attention · Panda Orb | ADD AUC: 58.86 -> 60.30 | Improved |
| Use window size [7,3,1] instead of [13,7,3] (ablation) | SGTAPose temporal cross-attention · Panda Orb | ADD AUC: 60.30 -> 57.11 | Hurt |
| Use window size [17,9,5] instead of [13,7,3] (ablation) | SGTAPose temporal cross-attention · Panda Orb | ADD AUC: 60.30 -> 56.54 | Hurt |
| Use window size [7,3,1] instead of [13,7,3] (ablation) | SGTAPose temporal cross-attention · Panda Orb | Median PCK (pix): 3.46 -> 3.41 | Mixed |
| Use temporal information and structure priors rather than single-frame estimation or crude temporal concatenation | camera-to-robot pose estimation under self-occlusion · Panda Orb severe self-occlusion | ADD AUC: 43.87 -> 59.37 | Improved |
| Use SGTAPose for camera-to-robot pose estimation in a static grasping task | robot grasping · Static grasping | R_object: 90.6% -> 100% | Improved |
| Use SGTAPose for camera-to-robot pose estimation in a dynamic grasping task | robot grasping · Dynamic grasping | R_object: 75.0% -> 93.8% | Improved |
Closest papers by meaning
- Markerless Camera-to-Robot Pose Estimation via Self-Supervised Sim-to-Real Transfer (CVPR 2023 · similarity 0.88)
- Real-time Holistic Robot Pose Estimation with Unknown States (ECCV 2024 · similarity 0.88)
- Single-View Robot Pose and Joint Angle Estimation via Render & Compare (CVPR 2021 · similarity 0.87)
- RoboTAG: End-to-end Robot Pose Estimation via Topological Alignment Graph (CVPR 2026 · similarity 0.87)
- RoboPEPP: Vision-Based Robot Pose and Joint Angle Estimation through Embedding Predictive Pre-Training (CVPR 2025 · similarity 0.86)
- PDF-HR: Pose Distance Fields for Humanoid Robots (NeurIPS 2026 · similarity 0.84)
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.
