AI Paper Atlas

Research map / Video / Action recognition

Alignment-guided Temporal Attention for Video Action Recognition

Yizhou Zhao, Zhenyang Li, Xun Guo, Yan Lu

NeurIPS 2022 · Video › Action recognition · approach: Spatiotemporal CNNs for action recognition · 9 citations (OpenAlex)

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 changedWhereMeasuredEffect
Add alignment-guided temporal attention to TimeSformerTimeSformer · Kinetics-400Top-1 accuracy: 81.1 -> 81.9Improved
Add alignment-guided temporal attention to TimeSformerTimeSformer · Something-Something V2Top-1 accuracy: 68.1 -> 67.0Mixed
Add de-alignment after the aligned temporal operation (ablation)ATA with CycleMLP-B5 · Kinetics-400Top-1 accuracy: 72.7 -> 77.7Improved
Add de-alignment after the aligned temporal operation (ablation)ATA with ConvNeXt-Base · Kinetics-400Top-1 accuracy: 76.0 -> 80.5Improved
Add de-alignment after the aligned temporal operation (ablation)ATA with ViT-Base · Kinetics-400Top-1 accuracy: 79.3 -> 79.6Improved
Replace temporal averaging with factorized temporal attention (ablation)CycleMLP-B5 · Kinetics-400Top-1 accuracy: 74.9 -> 76.8Improved
Replace temporal averaging with factorized temporal attention (ablation)ConvNeXt-Base · Kinetics-400Top-1 accuracy: 79.0 -> 80.1Improved
Replace temporal averaging with factorized temporal attention (ablation)ViT-Base · Kinetics-400Top-1 accuracy: 76.0 -> 78.0Improved
Add patch alignment to temporal attention (ablation)ViT-Base temporal attention · Kinetics-400Top-1 accuracy: 78.0 -> 79.6Improved
Add patch alignment to temporal attention (ablation)ViT-Base temporal attention · Something-Something V2Mutual information: 0.910 -> 1.290Improved
Insert ATA in one selected stage instead of no alignment (ablation)ViT-Base · Kinetics-400Top-1 accuracy: 78.0 -> 79.2; 78.0 -> 79.3; 78.0 -> 79.6Improved
Increase the number of ATA-equipped blocks (ablation)ViT-Base · Kinetics-400Top-1 accuracy: 79.3; 79.2; 79.4; 79.6Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy