Video Chat Emotion Detection

This Repo contains all files related to the creation of this blog post:

This is intended as a proof of concept for the live capture of emotional analytics from your audience while in a video meeting

Files

Dataset

• The file is too large to be uploaded to Github directly.
• Instead it can be downloaded here:
• https://www.kaggle.com/ashishpatel26/facial-expression-recognitionferchallenge

realtime_emotion.ipynb

• This is a colab notebook designed to be run on their free GPU's.
• In this notebook, I build, train, and export the model.

fer2.json & fer2.h5

• These are the saved model files for the verison I implement.
• You will need this if you plan to run the .py locally.

haarcascade_frontalface...xml

• Standard haar cascade file, used in the .py file.

emotion_live_capture.py

• This py file should be run from the terminal of the machine you want to screen capture.
• This will create a window and live analytics of the 7 emotions measured.
• Currently set to 4 max faces, you can change this in the file.
• I tuned the window locations for my machine with two monitors, so placement may be different on a different setup.
• This will output a comma separated text file of the captured results.

emotion_results.txt

• This is a sample report generated from running the .py file.

emotions_analytics

• This is a short notebook generating some reporting from the results.
• These images can be seen in the blog post mentioned above.

Acknowledgements:

• http://sefiks.com/2018/01/01/facial-expression-recognition-with-keras/
• https://medium.com/@jsflo.dev/training-a-tensorflow-model-to-recognize-emotions-a20c3bcd6468
• https://github.com/neha01/Realtime-Emotion-Detection
• https://github.com/makerportal/pylive