CERN: Compact facial expression recognition net

Darshan Gera,S. Balasubramanian, Anwesh Jami

Pattern Recognition Letters（2022）

引用 11|浏览12

暂无评分

摘要

•Build CERN, a low-calorie power-packed architecture for FER in the wild.•It has only 1.45M parameters, requires only 17MB and processes at real time 40 fps.•Superior performance over other light-weight architectures.•Comparable or superior performance over some of the heavy SOTA architectures.•Robust to occlusions and pose variations.

查看译文

关键词

Light-weight,Transfer learning,Facial expression recognition,Attention,Local-global context,CNN

AI 理解论文

溯源树

样例

生成溯源树，研究论文发展脉络

Chat Paper

正在生成论文摘要