Research map / Video / Action recognition
EAN: Event Adaptive Network for Enhanced Action Recognition
Yuan Tian, Yichao Yan, Guangtao Zhai, Guodong Guo, Zhiyong Gao
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 EABs to ResNet baseline (ablation) | ResNet baseline · Something-Something V1 (8F ablation) | top-1 accuracy (%): 48.6 -> 50.8 | Improved |
| add SOI-Tr to ResNet baseline (ablation) | ResNet baseline · Something-Something V1 (8F ablation) | top-1 accuracy (%): 48.6 -> 49.3 | Improved |
| add SOI-Tr on top of EABs (ablation) | ResNet+EABs · Something-Something V1 (8F ablation) | top-1 accuracy (%): 50.8 -> 51.9 | Improved |
| replace EAB stages with SOI-Tr (EAB stage 1~2, SOI-Tr 3~5) (ablation) | EAN 8F · Something-Something V1 (8F ablation) | top-1 accuracy (%): 51.9 -> 49.4 | Hurt |
| use EABs in all stages with no SOI-Tr (ablation) | EAN 8F · Something-Something V1 (8F ablation) | top-1 accuracy (%): 51.9 -> 50.8 | Hurt |
| use end-to-end Saliency-Net instead of Faster RCNN regions for SOI-Tr (ablation) | EAN 8F · Something-Something V1 (8F ablation) | top-1 accuracy (%): 51.1 -> 51.9 | Improved |
| use Faster RCNN regions instead of fixed regions (ablation) | SOI-Tr · Something-Something V1 (8F ablation) | top-1 accuracy (%): 50.9 -> 51.1 | No clear effect |
| dense interaction over all positions instead of sparse objects (ablation) | SOI-Tr · Something-Something V1 (8F ablation) | top-1 accuracy (%): 51.9 -> 51.3 | Hurt |
| larger dilated receptive field (L-Block vs S-Block) (ablation) | (2+1)D block · Something-Something V1 (8F ablation) | top-1 accuracy (%): 49.6 -> 50.3 | Improved |
| add multi-scale (Incep-Block vs L-Block) (ablation) | (2+1)D block · Something-Something V1 (8F ablation) | top-1 accuracy (%): 50.3 -> 50.8 | Improved |
| dynamic kernel fusion matrix instead of identity (Incep-Block) (ablation) | EAB · Something-Something V1 (8F ablation) | top-1 accuracy (%): 50.8 -> 51.9 | Improved |
| dynamic matrix instead of static matrix (ablation) | EAB · Something-Something V1 (8F ablation) | top-1 accuracy (%): 51.3 -> 51.9 | Improved |
Closest papers by meaning
- Action Recognition and Benchmark Using Event Cameras (TPAMI 2023 · similarity 0.89)
- Sympathy for the Details: Dense Trajectories and Hybrid Classification Architectures for Action Recognition (ECCV 2016 · similarity 0.88)
- Dynamic Normalization and Relay for Video Action Recognition (NeurIPS 2021 · similarity 0.88)
- VideoNet: A Large-Scale Dataset for Domain-Specific Action Recognition (CVPR 2026 · similarity 0.88)
- Action Recognition With Spatial-Temporal Discriminative Filter Banks (ICCV 2019 · similarity 0.88)
- Deep Alternative Neural Network: Exploring Contexts as Early as Possible for Action Recognition (NeurIPS 2016 · 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.
