Research map / Video / Action recognition
Dynamic Image Networks 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 |
|---|---|---|---|
| represent a video with a single dynamic image instead of its mean image | AlexNet video classification · UCF101 | accuracy: 52.6 -> 57.9 | Helped |
| represent a video with a single dynamic image instead of its max image | AlexNet video classification · UCF101 | accuracy: 45.4 -> 57.9 | Helped |
| replace Rank Pooling with Approx. Rank Pooling | rank pooling · ranking evaluation on 10 test videos | speed: 131 fps -> 5920 fps | Helped |
| replace Rank Pooling with Approx. Rank Pooling | rank pooling · ranking evaluation on 10 test videos | pairwise ranking accuracy: 99.5 ± 0.1 -> 96.5 ± 0.9 | Hurt |
| train dynamic images end-to-end instead of using fc6 max pooling and SVM (ablation) | MDI with AlexNet · HMDB51 | accuracy: 32.3 -> 35.8 | Helped |
| train dynamic images end-to-end instead of using fc6 max pooling and SVM (ablation) | MDI with AlexNet · UCF101 | accuracy: 68.6 -> 70.9 | Helped |
| apply RankPool after conv1 rather than pooling RGB frames into dynamic images (ablation) | AlexNet dynamic-map network · UCF101 | accuracy: 70.9 -> 67.1 | Hurt |
| combine dynamic-image features with static RGB frame features (ablation) | AlexNet video classification · HMDB51 | accuracy: 35.8 -> 42.8 | Helped |
| combine dynamic-image features with static RGB frame features (ablation) | AlexNet video classification · UCF101 | accuracy: 70.9 -> 76.9 | Helped |
| combine dynamic-image and static RGB features with trajectory features (ablation) | video action recognition · HMDB51 | accuracy: 60.0 -> 65.2 | Helped |
| combine dynamic-image and static RGB features with trajectory features (ablation) | video action recognition · UCF101 | accuracy: 86.0 -> 89.1 | Helped |
Closest papers by meaning
- Discriminative Hierarchical Rank Pooling for Activity Recognition (CVPR 2016 · similarity 0.90)
- Learning End-to-end Video Classification with Rank-Pooling (ICML 2016 · similarity 0.89)
- AdaScan: Adaptive Scan Pooling in Deep Convolutional Neural Networks for Human Action Recognition in Videos (CVPR 2017 · similarity 0.89)
- Improving Video Model Transfer With Dynamic Representation Learning (CVPR 2022 · similarity 0.89)
- Generalized Rank Pooling for Activity Recognition (CVPR 2017 · similarity 0.89)
- Dynamic Normalization and Relay for Video Action Recognition (NeurIPS 2021 · similarity 0.89)
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.