Research map / Video / Action recognition
Unsupervised Learning of View-invariant Action Representations
Read the paperCheck 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 view-adversarial training (ablation) | Fine-tuned encoder · NTU RGB+D cross-view | accuracy (%): 60.1 -> 63.9 | Helped |
| Add view-adversarial training (ablation) | Fine-tuned encoder · NTU RGB+D cross-subject | accuracy (%): 66.0 -> 68.1 | Helped |
| Add flow reconstruction loss (ablation) | Cross-view flow prediction framework · NTU RGB+D cross-subject, RGB input | cross-view flow prediction error: 0.0267 -> 0.0259 | Helped |
| Add view-adversarial training (ablation) | Cross-view flow prediction framework · NTU RGB+D cross-subject, RGB input | cross-view flow prediction error: 0.0259 -> 0.0254 | Helped |
| Add flow reconstruction loss (ablation) | Cross-view flow prediction framework · NTU RGB+D cross-view, RGB input | cross-view flow prediction error: 0.0265 -> 0.0252 | Helped |
| Add view-adversarial training (ablation) | Cross-view flow prediction framework · NTU RGB+D cross-view, RGB input | cross-view flow prediction error: 0.0252 -> 0.0248 | Helped |
| Fine-tune the unsupervised pretrained encoder rather than freeze it | Ours (depth) · NTU RGB+D cross-view | accuracy (%): 53.9 -> 63.9 | Helped |
| Use flow instead of depth input | Ours · NTU RGB+D cross-view | accuracy (%): 63.9 -> 83.4 | Helped |
| Use flow instead of RGB input | Ours · NTU RGB+D cross-view | cross-view flow prediction error: 0.0248 -> 0.0193 | Helped |
| Transfer unsupervised pretrained depth representation and fine-tune instead of train from scratch | Ours · MSR-DailyActivity3D cross-subject | accuracy (%): 42.5 -> 82.3 | Helped |
| Transfer unsupervised pretrained depth representation and fine-tune instead of train from scratch | Ours · Northwestern-UCLA cross-view | accuracy (%): 35.8 -> 62.5 | Helped |
| Train from scratch rather than use the unsupervised learned representation | Ours · NTU RGB+D cross-view, depth input | accuracy (%): 37.7 -> 63.9 | Hurt |
Closest papers by meaning
- Unsupervised Deep Learning of Mid-Level Video Representation for Action Recognition (AAAI 2018 · similarity 0.89)
- Enhancing Unsupervised Video Representation Learning by Decoupling the Scene and the Motion (AAAI 2021 · similarity 0.89)
- Unsupervised Representation Learning With Long-Term Dynamics for Skeleton Based Action Recognition (AAAI 2018 · similarity 0.89)
- Recognizing Actions in Videos From Unseen Viewpoints (CVPR 2021 · similarity 0.88)
- Exploiting Motion Information from Unlabeled Videos for Static Image Action Recognition (AAAI 2020 · similarity 0.88)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · 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.