Research map / Video / Action recognition
Making the Invisible Visible: Action Recognition Through Walls and Occlusions
Tianhong Li, Lijie Fan, Mingmin Zhao, Yingcheng Liu, Dina Katabi
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 spatio-temporal attention module (ablation) | RF-Action · RF-MMD (IoU 0.5) | mAP: 80.1 -> 87.8 | Improved |
| add spatio-temporal attention module (ablation) | RF-Action · PKU-MMD | mAP: 92.6 -> 92.9 and 94.2 -> 94.4 | Improved |
| add spatio-temporal attention module (ablation) | RF-Action · NTU-RGB+D | accuracy: 86.5 -> 86.8 and 91.1 -> 91.6 | Improved |
| add multi-proposal module instead of single proposal (ablation) | RF-Action · RF-MMD (IoU 0.5) | mAP: 65.5 -> 87.8 | Improved |
| add PKU-MMD vision skeleton data to RF-MMD training (multimodal training) (ablation) | RF-Action · RF-MMD test | mAP: 83.3 -> 87.8 | Improved |
| add RF-MMD to PKU-MMD training (multimodal training) (ablation) | RF-Action · PKU-MMD test | mAP: 92.9 -> 93.3 and 94.4 -> 94.9 | Improved |
| add PKU-MMD to training versus RF-MMD only, tested on PKU-MMD (ablation) | RF-Action · PKU-MMD test | mAP: 60.1 -> 93.3 and 60.4 -> 94.9 | Improved |
| train end-to-end so action loss backpropagates through skeleton generation (ablation) | RF-Action · RF-MMD (IoU 0.5) | mAP: 84.3 -> 87.8 | Improved |
| train end-to-end so action loss backpropagates through skeleton generation (ablation) | RF-Action · RF-MMD skeleton error | skeleton error (cm): 3.8 -> 3.4 | Improved |
| use camera ground-truth skeletons instead of RF-derived skeletons at inference | RF-Action · RF-MMD visible scenes (IoU 0.5) | mAP: 87.8 -> 90.5 | Improved |
| RF-Action versus HCN on the same skeletons | RF-Action · RF-MMD through-wall (IoU 0.5) | mAP: 75.9 -> 83.0 | Improved |
| occlusion by walls (through-wall versus visible) | RF-Action · RF-MMD (IoU 0.5) | mAP: 87.8 -> 83.0 | Hurt |
Closest papers by meaning
- View Adaptive Neural Networks for High Performance Skeleton-Based Human Action Recognition (TPAMI 2019 · similarity 0.87)
- InfoGCN: Representation Learning for Human Skeleton-Based Action Recognition (CVPR 2022 · similarity 0.87)
- Through-Wall Human Pose Estimation Using Radio Signals (CVPR 2018 · similarity 0.86)
- InfoGCN++: Learning Representation by Predicting the Future for Online Skeleton-Based Action Recognition (TPAMI 2024 · similarity 0.86)
- Graph Diffusion Convolutional Network for Skeleton Based Semantic Recognition of Two-Person Actions (TPAMI 2023 · similarity 0.86)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 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.
