Research map / 3D vision / Human pose estimation
AutoComPose: Automatic Generation of Pose Transition Descriptions for Composed Pose Retrieval Using Multimodal LLMs
Yi-Ting Shen, Sungmin Eum, Doheon Lee, Rohit Shete, Chiao-Yi Wang, Heesung Kwon, Shuvra S. Bhattacharyya
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 cyclic loss (ablation) | AutoComPose CLIP RN50 · FIXMYPOSE | R@1: 5.88 -> 8.24 | Improved |
| add swapping and mirroring augmentation (ablation) | AutoComPose CLIP RN50 · FIXMYPOSE | R@1: 2.35 -> 8.24 | Improved |
| add cyclic loss (ablation) | AutoComPose CLIP RN50 · PoseFixCPR | R@1: 56.48 -> 61.36 | Improved |
| add swapping and mirroring augmentation (ablation) | AutoComPose CLIP RN50 · PoseFixCPR | R@1: 48.15 -> 61.36 | Improved |
| add cyclic loss (ablation) | AutoComPose CLIP RN101 · FIXMYPOSE | R@1: 6.47 -> 6.47 | No clear effect |
| add cyclic loss (ablation) | AutoComPose CLIP RN101 · AIST-CPR | R@1: 5.25 -> 5.12 | No clear effect |
| add swapping and mirroring augmentation (ablation) | AutoComPose CLIP RN50 · AIST-CPR | R@50: 36.25 -> 50.75 | Improved |
| increase paraphrases from 1 to 3 per transition (ablation) | AutoComPose CLIP · FIXMYPOSE | R@1: 5.88 -> 8.24 | Improved |
| increase paraphrases from 3 to 5 per transition (ablation) | AutoComPose Combiner · FIXMYPOSE | R@1: 9.41 -> 8.82 | No clear effect |
| add cyclic training to rule-based descriptions (ablation) | rule-based CLIP RN50 · PoseFixCPR | R@1: 42.00 -> 55.05 | Improved |
| add paraphrasing (3 vs 1) to human descriptions (ablation) | human CLIP RN50 · PoseFixCPR | R@1: 48.73 -> 52.02 | Improved |
| add Stage I body part-based descriptions (ablation) | AutoComPose CLIP RN50 · FIXMYPOSE | R@1: 8.63 -> 8.24 | Mixed |
Closest papers by meaning
- PoseLLaVA: Pose Centric Multimodal LLM for Fine-Grained 3D Pose Manipulation (AAAI 2025 · similarity 0.86)
- MotionMap: Representing Multimodality in Human Pose Forecasting (CVPR 2025 · similarity 0.85)
- FlexPose: Pose Distribution Adaptation with Limited Guidance (AAAI 2025 · similarity 0.85)
- UniPose: A Unified Multimodal Framework for Human Pose Comprehension, Generation and Editing (CVPR 2025 · similarity 0.85)
- HiPoser: 3D Human Pose Estimation with Hierarchical Shared Learning at Parts-Level Using Inertial Measurement Units (AAAI 2025 · similarity 0.85)
- ChatPose: Chatting about 3D Human Pose (CVPR 2024 · similarity 0.85)
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.
