AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

ReactioNet: Learning High-Order Facial Behavior from Universal Stimulus-Reaction by Dyadic Relation Reasoning

Xiaotian Li, Taoyue Wang, Geran Zhao, Xiang Zhang, Xi Kang, Lijun Yin

ICCV 2023 · Graph learning › Knowledge and temporal graphs · approach: Graph reasoning for visual recognition · 6 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 cross-domain contrastive learning (CDC) (ablation)DRR · ReactioNet AUDF1: 58.9 -> 61.3Improved
add homophily relation module (HO) (ablation)DRR · ReactioNet AUDF1: 59.1 -> 61.3Improved
add heterophily relation module (HE) (ablation)DRR · ReactioNet scene classificationaccuracy: 56.1 -> 60.2Improved
add heterophily relation module (HE) (ablation)DRR · ReactioNet facial expression recognitionaccuracy: 73.5 -> 78.5Improved
add stimulus and text domain information (DRR dynamic vs ResNet18)ResNet-18 · ReactioNet AUDF1: 54.8 -> 61.3Improved
feature-fusion of reaction and stimulus without relation reasoningFeature fusion baseline · ReactioNet AUDF1: 54.8 -> 54.6No clear effect
use temporal sequential frames (DRR dynamic) instead of single frame (DRR static)DRR · ReactioNet facial expression recognitionaccuracy: 76.2 -> 78.5Improved
multi-task dyadic reasoning (DRR dyadic) instead of dynamic on AUDDRR · ReactioNet AUDF1: 61.3 -> 61.1No clear effect
apply DRR static to film and sports scenesDRR Static · ReactioNet AUD by sceneF1: ResNet-18 53.9 -> 50.4 on film sceneHurt
pretrain on ReactioNet (DRR PR) and fine-tune on benchmark datasetsvisual encoder · RAFDB FERaccuracy: 79.1 -> 89.6 (vs baseline GCN)Improved
DRR variants on ten benchmark datasetsDRR · 5 AUD and 5 FER datasetsF1 or accuracy: best on seven of tenMixed
train on ReactioNet and test on other datasets versus the reverseResNet-18 · cross-database BP4D, DISFA, EmotioNet, AffectNetF1 and accuracy: up to 23%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