AI Paper Atlas

Research map / Video / Action recognition

Exploiting Feature and Class Relationships in Video Categorization with Regularized Deep Neural Networks

Yu–Gang Jiang, Zuxuan Wu, Jun Wang, Xiangyang Xue, Shih‐Fu Chang

TPAMI 2017 · Video › Action recognition · approach: Contrastive video representation learning · 415 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 fusion-layer feature-relationship regularization (ablation)DNN · FCVIDmAP: 72.1% -> 75.4%Improved
add fusion-layer feature-relationship regularization (ablation)DNN · Hollywood2mAP: 64.2% -> 65.9%Improved
add class-relationship regularization on output layer (ablation)DNN · FCVIDmAP: 72.1% -> 74.4%Improved
add class-relationship regularization on output layer (ablation)DNN · Hollywood2mAP: 64.2% -> 65.1%Improved
add both regularizations (full rDNN) over rDNN-F (ablation)rDNN-F · FCVIDmAP: 75.4% -> 76.0%Improved
add both regularizations (full rDNN) over rDNN-C (ablation)rDNN-C · Hollywood2mAP: 65.1% -> 66.9%Improved
learn class relationship matrix V instead of fixing V to identity (ablation)rDNN-C · Hollywood2, CCV, FCVIDmAP: drops 1.6, 0.8 and 3.0 percentImproved
initialize V with prior class knowledge (ablation)rDNN-C · FCVIDmAP: 74.4% -> 75.0%No clear effect
deepen plain DNN to six layers (ablation)DNN · FCVIDmAP: 72.1% -> 67.3%Hurt
deepen plain DNN to eight layers (ablation)DNN · Hollywood2mAP: 64.2% -> 56.2%Hurt
use rDNN-C instead of DMF post-hoc class refinement (ablation)DNN · Hollywood2mAP: 61.8% -> 65.1%Improved
use rDNN-F instead of SVM early fusion (ablation)SVM-EF · Hollywood2mAP: 64.1% -> 65.9%Improved

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