Research map / 3D vision / Human pose estimation
"PoseEmbroider: Towards a 3D, Visual, Semantic-aware Human Pose Representation"
Ginger Delmas, Philippe Weinzaepfel, Francesc Moreno-Noguer, Gregory Rogez
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 |
|---|---|---|---|
| replace aligner with PoseEmbroider (transformer with unimodal reprojected contrastive losses) | multi-modal pose retrieval · BEDLAM-Script val | total mRecall: 72.5 -> 74.6 | Improved |
| train on both single and dual input sets instead of single inputs only (ablation) | PoseEmbroider · BEDLAM-Script val | total mRecall: 69.7 -> 74.6 | Improved |
| add dual-input training to the Aligner (ablation) | Aligner · BEDLAM-Script val | total mRecall: 72.4 -> 72.5 | No clear effect |
| use transformer core instead of MLP core (ablation) | PoseEmbroider · BEDLAM-Script val | total mRecall: 73.4 -> 74.6 | Improved |
| add modality re-projection heads with unimodal contrastive losses (ablation) | PoseEmbroider · BEDLAM-Script val | total mRecall: 73.5 -> 74.6 | Improved |
| use PoseEmbroider instead of Aligner as pose encoder for instruction generation | pose instruction generation · BEDLAM-Fix (PA,PB) | R@1: 31.7 -> 43.1 | Improved |
| use PoseEmbroider instead of Aligner for instruction generation | pose instruction generation · PoseFix-OOS test | R@1: 19.5 -> 27.4 | Improved |
| use PoseEmbroider instead of Aligner for image-pair instruction generation | pose instruction generation · BEDLAM-Fix (VA,VB) | BLEU-4: 27.8 -> 27.4 | Mixed |
| add text hint to image input for SMPL regression | PoseEmbroider · BEDLAM-Script val | pa-MPJPE (mm): 49 -> 44 | Improved |
| use PoseEmbroider instead of Aligner features for SMPL regression | SMPL regression · 3DPW test | pa-MPJPE (mm): 54 -> 53 | No clear effect |
Closest papers by meaning
- ChatPose: Chatting about 3D Human Pose (CVPR 2024 · similarity 0.89)
- UniPose: A Unified Multimodal Framework for Human Pose Comprehension, Generation and Editing (CVPR 2025 · similarity 0.89)
- Stable-Pose: Leveraging Transformers for Pose-Guided Text-to-Image Generation (NeurIPS 2024 · similarity 0.87)
- Tri-Modal Motion Retrieval by Learning a Joint Embedding Space (CVPR 2024 · similarity 0.87)
- PoseLLaVA: Pose Centric Multimodal LLM for Fine-Grained 3D Pose Manipulation (AAAI 2025 · similarity 0.87)
- Semantics-aware Test-time Adaptation for 3D Human Pose Estimation (ICML 2025 · similarity 0.87)
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.
