CS475_facial expression recognition

This project uses tensorflow.

For Apple M1 and newest chip carried MacBook Pro, refer to env_apple_M1.sh

If you find an error message that a certain package is not found, you will need to install the package and contact either one of us at: xzhao77@jhu.edu, ylu106@jhu.edu, zsha2@jhu.edu, pxu11@jhu.edu, if you still have troubles.

Here are the instructions to run the facial expression recognition application in src.

python src/facial_recog_camera.py -i <input video directory/ 0 for camera> -o <output file directory (preferred .avi format)>

example:

python src/facial_recog_camera.py -i data/test_video.mp4 -o output/out.avi

python src.facial_recog_camera.py -i 0 -o output/cameraout.avi

Here are the instructions to train model in src.

python src/train.py --dataset <#name of dataset> --model <#name of model> --epoches <#number of epochs> --batch-size <#batch size>

Note that the <#name of dataset> is either combined, raf, or ck+.

Note also that <#name of model> is either SimpleCNN3, GaborCNN3,or VGG16.

Data set

Since we cannot copy, publish or distribute any portion of the RAF database. We remove the RAF data from this github repo.

Results

Please see graphs below and output/curves.ipynb for our model results.

ck+

combined

Please see below for our sample video results.

video output

camera output