Research map / Video / Action recognition
Recognizing Actions in Videos From Unseen Viewpoints
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the NPL with multi-view projections | ResNet-50 + Kinetics · Human3.6M unseen | accuracy: 38.2% -> 87.5% | Helped |
| Add the NPL with multi-view projections | ResNet-50 + Kinetics · MLB unseen | accuracy: 30.2% -> 42.7% | Helped |
| Add the NPL with multi-view projections | ResNet-50 + Kinetics · Toyota SmartHome unseen | accuracy: 34.2% -> 39.6% | Helped |
| Pretrain on Kinetics | ResNet-50 · Human3.6M unseen | accuracy: 9.1% -> 38.2% | Helped |
| Pretrain on Kinetics | ResNet-50 · MLB unseen | accuracy: 27.3% -> 30.2% | Helped |
| Pretrain on Kinetics | ResNet-50 · Toyota SmartHome unseen | accuracy: 33.7% -> 34.2% | No clear effect |
| Use 3D pose features instead of the video representation | Action recognition model · Toyota SmartHome unseen | accuracy: 34.2% -> 11.6% | Hurt |
| Use 3D pose features instead of the video representation | Action recognition model · MLB seen and unseen | accuracy: 55.6% -> 36.5%; 30.2% -> 33.6% | Mixed |
| Add multi-view projections without geometric loss constraints (ablation) | NPL · MLB unseen | accuracy: 30.2 -> 28.5 | Hurt |
| Add multi-view projections without geometric loss constraints (ablation) | NPL · Toyota SmartHome unseen | accuracy: 34.2 -> 33.7 | Hurt |
| Add 3D representation loss with multi-view projections (ablation) | NPL · MLB unseen | accuracy: 28.5 -> 35.5 | Helped |
| Add 3D representation loss with multi-view projections (ablation) | NPL · Toyota SmartHome unseen | accuracy: 33.7 -> 37.8 | Helped |
Closest papers by meaning
- 3D Action Recognition From Novel Viewpoints (CVPR 2016 · similarity 0.89)
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.88)
- Unsupervised Learning of View-invariant Action Representations (NeurIPS 2018 · similarity 0.88)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.88)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.87)
- Viewpoint Rosetta Stone: Unlocking Unpaired Ego-Exo Videos for View-invariant Representation Learning (CVPR 2025 · 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.