Research map / 3D vision / Human pose estimation
Learning Human Pose Models from Synthesized Data for Robust RGB-D Action Recognition
Jian Liu, Hossein Rahmani, Naveed Akhtar, Ajmal Saeed Mian
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 |
|---|---|---|---|
| refine synthetic RGB images with SimGAN (ablation) | HPM_RGB · UWA3D Multiview-II | mean accuracy: 64.7 -> 68.0 | Improved |
| refine synthetic RGB images with SimGAN (ablation) | HPM_RGB · N-UCLA Multiview | mean accuracy: 76.4 -> 77.8 | Improved |
| refine synthetic depth images with GAN (ablation) | HPM_3D · UWA3D Multiview-II | mean accuracy: 73.8 -> 74.8 | Improved |
| refine synthetic depth images with GAN (ablation) | HPM_3D · N-UCLA Multiview | mean accuracy: 78.4 -> 79.7 | Improved |
| use proposed synthetic data instead of SURREAL (ablation) | HPM_RGB · UWA3D Multiview-II | mean accuracy: 67.4 -> 68.0 | Improved |
| use proposed synthetic data instead of SURREAL (ablation) | HPM_3D · UWA3D Multiview-II | mean accuracy: 72.1 -> 74.8 | Improved |
| use proposed synthetic data instead of SURREAL (ablation) | HPM_RGB · N-UCLA Multiview | mean accuracy: 74.4 -> 77.8 | Improved |
| fine-tune GoogLeNet on synthetic pose data (ablation) | HPM_Orig · UWA3D Multiview-II | mean accuracy: 62.8 -> 68.0 | Improved |
| fine-tune GoogLeNet on synthetic pose data (ablation) | HPM_Orig · N-UCLA Multiview | mean accuracy: 66.7 -> 77.8 | Improved |
| add dense trajectory features | HPM_RGB · UWA3D Multiview-II | mean accuracy: 68.0 -> 76.4 | Improved |
| add dense trajectory features | HPM_RGB · NTU RGB+D cross-view | accuracy: 72.9 -> 83.2 | Improved |
| fine-tune baseline with the synthetic data | C3D · UWA3D Multiview-II | accuracy: +2.3 | Improved |
Closest papers by meaning
- Synthetic Humans for Action Recognition from Unseen Viewpoints (IJCV 2021 · similarity 0.90)
- 3D Action Recognition From Novel Viewpoints (CVPR 2016 · similarity 0.90)
- Learning Action Recognition Model From Depth and Skeleton Videos (ICCV 2017 · similarity 0.89)
- Multi-task Deep Learning for Real-Time 3D Human Pose Estimation and Action Recognition (TPAMI 2020 · similarity 0.89)
- Synthesizing Images of Humans in Unseen Poses (CVPR 2018 · similarity 0.88)
- Just Add ?! Pose Induced Video Transformers for Understanding Activities of Daily Living (CVPR 2024 · 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.
