Facial Expression Recognition by Convolutional Neural Network

Final Project for Deep Learning (Fall 2018)@ GWU
Dec. 2018

Authors: Xiaochi Li, Liwei Zhu, Jia Chen

Best model: 74% Accuracy Download the trained model
Read our presentation

Data Source:

Real-world Affective Faces Database (RAF-DB): http://www.whdeng.cn/RAF/model1.html
To show respect to the owner, the RAF-DB data has been removed from this repo.

Usage (Training Neural Network):

Enviroment requirement: Tensorflow 1.10.0, Keras 2.2.2

0. Contact the owner of RAF-DB to get the data set. And put aligned.zip and list_patition_label.txt under /Code/RAFDB/.
1. Unzip the data: unzip /Code/RAFDB/aligned.zip
2. Change main program to executable chmod u+x /Code/Prototype/V2_main.py
3. Execute main program /Code/Prototype/V2_main.py or python3 /Code/Prototype/V2_main.py
   For MLP: python3 /Code/Prototype/MLP_main.py
4. Type the name of your run in python console
5. Find the plot in /Code/Prototype/imgs/, the saved model in /Code/Prototype/trained_model, the tensorboard log in /Code/Prototype/logs/.
   For MLP: Find the plot in /Code/Prototype/imgs_C/, the saved model in /Code/Prototype/trained_model_C, the tensorboard log in /Code/Prototype/logs_MLP/

Place to modify:
In /Code/Prototype/V2_main.py: Line 20 model = xxx choose the model you want to see
In /Code/Prototype/MLP_main.py: Line 20 model = xxx choose the model you want to see

Usage (Real time facial expression detection)

Environment requirement: opencv 3.4.3, Tensorflow 1.10.0, Keras 2.2.2

Execute main program python3 Code/real_time_detection/RTD_V2.py

Place to modify:
In /Code/real_time_detection/RTD_helper.py Line 24, 25 Choose the model you want to use