FER - Facial Expression Recognition

Opensource deep learning framework TensorFlow is used in Facial Expression Recognition(FER). The trained models achieved 65% accuracy in fer2013. If you like this, please give me a star.

FER requires:

• Python (>= 3.3)
• TensorFlow (>= 1.1.0) Installation
• OpenCV (python3-version) Installation

To run the demo, just type:

python3 main.py

Then the program will create a window to display the scene capture by webcamera. You need press SPACE key to capture face in current frame and recognize the facial expression.

If you just want to run this demo instead of training the model from scratch, the following content can be skipped.

If you want to train a model from scratch by yourself, download the fer2013 datasets in kaggle(91.97MB). Then extract the data to data/fer2013 folder.

It's is import that modifying the MODE(in main.py) from demo to train before you start training. Then type:

python3 main.py