Research map / Video / Action recognition
Multi-view Action Recognition using Cross-view Video Prediction
Shruti Vyas, Yogesh S Rawat, Mubarak Shah
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 cross-view video prediction (versus no prediction) (ablation) | RL-NET action recognizer · NTU-RGB+D cross-subject | accuracy: 63.0 -> 88.9 | Improved |
| add cross-view video prediction (versus no prediction) (ablation) | RL-NET action recognizer · NTU-RGB+D cross-view | accuracy: 65.2 -> 86.3 | Improved |
| predict an unseen view instead of a seen view (ablation) | RL-NET · NTU-RGB+D cross-subject | accuracy: 71.4 -> 88.9 | Improved |
| predict an unseen view instead of a seen view (ablation) | RL-NET · NTU-RGB+D cross-view | accuracy: 78.1 -> 86.3 | Improved |
| add seen-view prediction (versus no prediction) (ablation) | RL-NET · NTU-RGB+D cross-subject | accuracy: 63.0 -> 71.4 | Improved |
| randomly select input views (M-1) instead of fixed input views (M-2) (ablation) | rendering training · NTU-RGB+D cross-subject | average accuracy: 59.3 -> 74.6 | Improved |
| render an unseen view (M-1) instead of a seen view at random time (M-3) (ablation) | rendering training · NTU-RGB+D cross-subject | average accuracy: 56.3 -> 74.6 | Improved |
| use all views at test time instead of a single view (ablation) | Proposed · NTU-RGB+D RGB cross-subject | accuracy: 82.3 -> 88.9 | Improved |
| use all views at test time instead of a single view (ablation) | Proposed · NTU-RGB+D depth cross-subject | accuracy: 71.8 -> 79.4 | Improved |
| increase test clips from 6 to full video length (ablation) | Proposed · NTU-RGB+D cross-subject | accuracy: 74.6 -> 82.3 | Improved |
| transfer representation from NTU-RGB+D instead of training from scratch (ablation) | Proposed · N-UCLA cross-subject | accuracy: 35.1 -> 87.5 | Improved |
| transfer representation from NTU-RGB+D instead of training from scratch (ablation) | Proposed · N-UCLA cross-view | accuracy: 43.4 -> 83.1 | Improved |
Closest papers by meaning
- Unsupervised Learning of View-invariant Action Representations (NeurIPS 2018 · similarity 0.90)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.90)
- Multi-Modal Multi-Action Video Recognition (ICCV 2021 · similarity 0.89)
- Recognizing Actions in Videos From Unseen Viewpoints (CVPR 2021 · similarity 0.89)
- Dividing and Aggregating Network for Multi-view Action Recognition (ECCV 2018 · similarity 0.88)
- MMNet: A Model-Based Multimodal Network for Human Action Recognition in RGB-D Videos (TPAMI 2022 · similarity 0.88)
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.
