AI Paper Atlas

Research map / Video / Action recognition

HACS: Human Action Clips and Segments Dataset for Recognition and Temporal Localization

Hang Zhao, Antonio Torralba, Lorenzo Torresani, Zhicheng Yan

ICCV 2019 · Video › Action recognition · approach: Disentangled action recognition · 293 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
pretrain on HACS Clips instead of Kinetics-600 (RGB)I3D · Kinetics-400 valVideo@1 accuracy: 72.9 -> 73.4Improved
pretrain on HACS Clips instead of Kinetics-600 (RGB)I3D · HMDB51Video@1 accuracy: 73.4 -> 73.6Improved
pretrain on HACS Clips instead of Kinetics-600 (Flow)I3D · UCF101Video@1 accuracy: 96.0 -> 95.7Hurt
pretrain on HACS Clips instead of no pretraining (RGB)I3D · HMDB51Video@1 accuracy: 39.4 -> 73.6Improved
pretrain on HACS Clips instead of no pretraining (RGB)I3D · UCF101Video@1 accuracy: 75.0 -> 95.1Improved
add optical flow to RGB with late fusion (ablation)I3D · HACS Clips valClass@1: 80.3 -> 83.5Improved
sample clips with maximum consensus instead of random for trainingI3D/Res3D · Validation setsClass@1Improved
maximum consensus sampling instead of random samplingclip sampling · Train-mini clipsnumber of positive clips: 82.2K -> 100.3KImproved
maximum entropy sampling instead of random samplingclip sampling · Train-mini clipsnumber of positive clips: 82.2K -> 71.3KHurt
train BSN on full HACS Segments instead of Mini (ablation)BSN · HACS SegmentsAR@100: 61.85 -> 63.62Improved
train SSN on full HACS Segments instead of Mini (ablation)SSN · HACS Segmentsaverage mAP: 15.93 -> 18.97Improved
evaluate on HACS Segments Mini instead of ActivityNetBSN · localization difficultyAR@100: 74.16 -> 61.85Hurt

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