AI Paper Atlas

Research map / Video / Action recognition

Selective Volume Mixup for Video Action Recognition

Yi Tan, Zhaofan Qiu, Yanbin Hao, Ting Yao, Tao Mei

IJCV 2026 · Video › Action recognition · approach: Contrastive video representation learning · 5 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 SV-Mix augmentationTSM · Sth-Sth V1top-1 accuracy (%): 45.5 -> 47.2Improved
add SV-Mix augmentationViViT · Diving48top-1 accuracy (%): 70.0 -> 76.2Improved
add SV-Mix augmentationUniformer · UCF101top-1 accuracy (%): 93.2 -> 97.1Improved
add SV-Mix augmentationMViTv2 · Mini-Kineticstop-1 accuracy (%): 79.3 -> 79.5Improved
use Mixup augmentation on videoTSM · Sth-Sth V1top-1 accuracy (%): 45.5 -> 44.6Hurt
replace Cutmix/Mixup with SV-MixTSM · UCF101top-1 accuracy (%): 86.9 -> 88.4Improved
add spatial selective module over Cutmix (ablation)TSM · Sth-Sth V1top-1 accuracy (%): 45.7 -> 47.0Improved
add temporal selective module over Mixup (ablation)TSM · Sth-Sth V1top-1 accuracy (%): 44.6 -> 46.6Improved
combine temporal and spatial modules by random switching (ablation)SV-Mix · Sth-Sth V1top-1 accuracy (%): 47.0 -> 47.2Improved
use disentangled (EMA target) training instead of entangled training (ablation)SV-Mix spatial selective · Sth-Sth V1top-1 accuracy (%): 45.6 -> 47.0Improved
use disentangled training instead of entangled training (ablation)SV-Mix both modules · Sth-Sth V1top-1 accuracy (%): 45.8 -> 47.2Improved
add lambda embedding (ablation)SV-Mix · Sth-Sth V1top-1 accuracy (%): 45.1 -> 46.9Improved

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