AI Paper Atlas

Research map / Video / Action recognition

Mitigating and Evaluating Static Bias of Action Representations in the Background and the Foreground

Haoxin Li, Yuan Liu, Hanwang Zhang, Boyang Li

ICCV 2023 · Video › Action recognition · approach: Disentangled action recognition · 22 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 StillMix augmentationTSM · Kinetics-400 SCUFOtop-1 accuracy (lower is better): 17.22 -> 5.23Improved
Add StillMix augmentationTSM · Kinetics-400 IIDtop-1 accuracy: 71.13 -> 71.28No clear effect
Use ActorCutMix or FAME background replacementTSM, Swin-T · SCUBA vs SCUFOSCUBA and SCUFO accuracy: SCUBA 37.39 -> 40.91 (FAME), SCUFO 17.22 -> 18.34Mixed
Use VideoMix, SDN or BETSM, Swin-T · Kinetics-400 SCUBAtop-1 accuracy: 37.39 -> 36.95 (SDN, TSM)No clear effect
Choose biased frames via reference network instead of uniform sampling (ablation)StillMix · HMDB51 contrasted accuracy, TSMcontrasted accuracy: 31.21 -> 33.14Improved
Sample low-probability frames from reference network instead of high-probability (ablation)StillMix · HMDB51 contrasted accuracy, TSMcontrasted accuracy: 33.14 -> 29.17Hurt
Use reference network of the same architecture as main network (ablation)StillMix · UCF101 contrasted accuracy, Swin-Tcontrasted accuracy: 31.09 -> 32.14Improved
Mix action labels in StillMix (ablation)StillMix · UCF101/HMDB51performanceHurt
Use StillMix pretraining features for weakly supervised localizationBaSNet on TSM RGB features · THUMOS14avg mAP: 0.1810 -> 0.1935Improved
Add StillMix on conflicting-foreground videosSwin-T · Kinetics-400 binary conflicting FGaccuracy: 72.83 -> 85.51Improved

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