I used this project to learn the basics of converting a Keras model trained in Python to a JavaScript model
Implement an emotion-classification model directly in browser to detect a person's current emotion via the webcam, and display the detected emotion to the user
This project is implemented purely in JavaScript using TensorFlowJS. The facial-recognition model is ported from blazeface and the emotion-detection model was trained in Python on the fer-2013 dataset
| Angry | Surprise |
|---|---|
 |
 |
- Extract bounding box for face from output
- Port classification code to looping function
- Center webcam stream
- Show percentage emotion using output tensor
- Incorporate Bulma
- Transition between colour changes
- Push to heroku