Contains code to use a VGG model trained AffectNet to recognize four primary emotional classes (happiness, sadness, anger, neutral).
The code was developed using Tensorflow, Keras and OpenCV.
- Tensorflow 1.14.0
- Keras 2.3.0
- OpenCV 4.1.1
- python 3.7.4
Downloading the model is required before running the code. Please download the model (https://utdallas.box.com/s/3b16v2yygj73hx31tjs5w1wqqbmlqt21) and place it in the output directory.
The demo is in a .py and jupyter notebook. Emo_Detect contains the demo. Simply modify the glob() function to point to the image files. When the program is done running a results.csv will be generated in the same directory.