AI Paper Atlas

Research map / Video / Action recognition

Skim then Focus: Integrating Contextual and Fine-grained Views for Repetitive Action Counting

Zhengqi Zhao, Xiaohu Huang, Hao Zhou, Kun Yao, Errui Ding, Jingdong Wang, Xinggang Wang, Wenyu Liu, Bin Feng

IJCV 2025 · Video › Action recognition · approach: Fully supervised temporal action detection · 7 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 skim branch (ablation)focus branch only · RepCount Part-AMAE / OBO: 0.3161 -> 0.2794; 0.3709 -> 0.4238Improved
add LSAG (ablation)focus branch only · RepCount Part-AMAE / OBO: 0.3161 -> 0.2901; 0.3709 -> 0.3974Improved
add both skim branch and LSAG (ablation)focus branch only · RepCount Part-AMAE / OBO: 0.3161 -> 0.2489; 0.3709 -> 0.5166Improved
add skim branch (ablation)SkimFocusNet · Multi-RepCountMAE / OBO: 0.6014 -> 0.3514; 0.2133 -> 0.3400Improved
use top N_C sampling instead of random sampling (ablation)SkimFocusNet · RepCount Part-AMAE / OBO: 0.2570 -> 0.2489; 0.4503 -> 0.5166Improved
use uniform sampling instead of top N_C (ablation)SkimFocusNet · RepCount Part-AOBO: 0.5166 -> 0.4636Hurt
add feature adaption block in LSAG (ablation)LSAG · RepCount Part-AMAE / OBO: 0.2559 -> 0.2489; 0.4437 -> 0.5166Improved
add long-short relation modeling (ablation)LSAG · RepCount Part-AMAE / OBO: 0.3111 -> 0.2489; 0.4040 -> 0.5166Improved
use VideoSwin encoder instead of ResNet (ablation)dual-branch with density-map decoder · RepCount Part-AMAE: 0.5339 -> 0.2489Improved
use density-map decoder instead of classification (ablation)dual-branch with VideoSwin · RepCount Part-AMAE: 0.2689 -> 0.2489Improved
increase skim frames N_C from 32 to 128 (ablation)SkimFocusNet · RepCount Part-AMAE / OBO: 0.2489 -> 0.2670; 0.5166 -> 0.4702Hurt
increase long-short blocks B from 3 to 5 (ablation)LSAG · RepCount Part-AMAE / OBO: 0.2489 -> 0.2828; 0.5166 -> 0.4437Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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