AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Incremental Boosting Convolutional Neural Network for Facial Action Unit Recognition

Shizhong Han, Zibo Meng, AHMED-SHEHAB KHAN, Yan Tong

NeurIPS 2016 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis · 89 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
replace CNN with IB-CNNIB-CNN · SEMAINEaverage F1: 0.347 -> 0.416Improved
replace CNN with IB-CNNIB-CNN · SEMAINEaverage 2AFC: 0.735 -> 0.775Improved
replace boosting CNN (B-CNN) with IB-CNN, i.e. add incremental accumulation over timeIB-CNN · SEMAINEaverage F1: 0.310 -> 0.416Improved
use only strong-classifier loss (IB-CNN-S) instead of strong plus weak losses (ablation)IB-CNN · SEMAINEaverage F1: 0.412 -> 0.416Improved
IB-CNN on the rarest AU28 versus CNN and B-CNNIB-CNN · SEMAINE AU28F1: 0.280 -> 0.490Improved
plain B-CNN versus CNN on infrequent AUsB-CNN · SEMAINEaverage F1: 0.347 -> 0.310Hurt
replace CNN with IB-CNNIB-CNN · DISFAaverage F1: 0.405 -> 0.457Improved
replace B-CNN with IB-CNNIB-CNN · DISFAaverage 2AFC: 0.815 -> 0.823Improved
IB-CNN on infrequent AU5 versus CNNIB-CNN · DISFA AU5F1: 0.195 -> 0.312Improved
IB-CNN on AU15 versus CNNIB-CNN · DISFA AU15F1: 0.174 -> 0.135Hurt
vary eta scale c from 0.5 to 16 (ablation)IB-CNN · SEMAINEaverage F1No clear effect
vary number of IB-layer input neurons 64 to 2048 (ablation)IB-CNN · SEMAINEF1No clear effect

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