/FAER

This is the repository for project FAER name: Facial attribute estimation and expression recognition based on contextual channel attention mechanism

Primary LanguagePythonMIT LicenseMIT

基于上下文通道注意力机制的人脸属性估计与表情识别

#（Facial attribute estimation and expression recognition based on contextual channel attention mechanism）

overview

A PyTorch implementation of the FAER.

Training

We provide the training code for AffectNet and RAF-DB.

For AffectNet、RAF-DB dataset, run:

Comming soon

For CelebA dataset, run:

CUDA_VISIBLE_DEVICES=0,1 python main_celeba.py

Models

Pre-trained models can be downloaded for evaluation as following:

task accuracy link
CelebA 91.87% Comming soon
AffectNet 66.66% Comming soon
RAF-DB 91.75% Comming soon

Grad CAM++ Reproduction

overview

overview

For more detail imformation, see:

基于上下文通道注意力机制的人脸属性估计与表情识别