Research map / Video / Action recognition
VLAD3: Encoding Dynamics of Deep Features for Action Recognition
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 |
|---|---|---|---|
| Use VLAD3 instead of T3 temporal-pyramid pooling | C3D action recognition · UCF101 | accuracy: 82.96 -> 84.08 | Helped |
| Use VLAD3 instead of T3 temporal-pyramid pooling | C3D action recognition · Olympic Sports | mAP: 80.14 -> 90.78 | Helped |
| Use VLAD3 instead of T3 temporal-pyramid pooling | C3D action recognition · THUMOS15 | mAP: 56.50 -> 76.84 | Helped |
| Fuse VLAD3 with holistic average-pooled CNN features | VLAD3 · UCF101 | accuracy: 84.08 -> 84.65 | Helped |
| Fuse VLAD3 with holistic average-pooled CNN features | VLAD3 · Olympic Sports | mAP: 90.78 -> 90.81 | Helped |
| Fuse VLAD3 with holistic average-pooled CNN features | VLAD3 · THUMOS15 | mAP: 76.84 -> 78.15 | Helped |
| Use a single LDS over the full video rather than LDSs on subsequences pooled by VLAD3 (ablation) | VLAD3 · Olympic Sports and THUMOS15 | mAP | Hurt |
| Increase subsequence temporal support τ from short windows toward the support of the video dynamics (ablation) | VLAD3 · UCF101, Olympic Sports, and THUMOS15 | mAP or accuracy | Helped |
| Extend subsequence temporal support beyond the point where dynamics become heterogeneous (ablation) | VLAD3 · UCF101, Olympic Sports, and THUMOS15 | mAP or accuracy | Hurt |
| Use a single LDS for full-video dynamics instead of VLAD3 (ablation) | VLAD3 · UCF101 | accuracy | No clear effect |
| Use VLAD3 instead of T3 on UCF101 | C3D action recognition · UCF101 | accuracy: 82.96 -> 84.08 | Helped |
Closest papers by meaning
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.88)
- Dynamic Normalization and Relay for Video Action Recognition (NeurIPS 2021 · similarity 0.87)
- Recognize Actions by Disentangling Components of Dynamics (CVPR 2018 · similarity 0.87)
- Unsupervised Learning of View-invariant Action Representations (NeurIPS 2018 · similarity 0.86)
- More Is Less: Learning Efficient Video Representations by Big-Little Network and Depthwise Temporal Aggregation (NeurIPS 2019 · similarity 0.86)
- Improving Video Model Transfer With Dynamic Representation Learning (CVPR 2022 · 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.