Conversational agents assignment: math tutoring chatbot with anger management
The user's emotion can be detected through an external Python script. This script uses an older
version of tensorflow: v1.8.0, which is only available via conda package manager (not pip). When
emotion_recognition.py is run it sets up a local API on http://127.0.0.1:5000/. Add
get_emotion to the URL to do a GET request from furhat to retrieve the valence and
arousal values of the user trough a webcam image. Don't forget to slide the little cover off your
webcam as I always do ;).
val response = khttp.get("http://127.0.0.1:5000/get_emotion").text
val values = response.split(' ')
val valence = values[0].toDouble()
val arousal = values[1].toDouble()
Install the python dependencies
pip install -r requirements.txt
Copy the weights from https://drive.google.com/drive/folders/1zQ5z9v1abgfw61R6Jcy3PGrJ3IZ9XVMX to the scripts/resources/ folder.
Run this to start the server:
python scripts/emotion_api.py
And allow all the necessary microphone and camera access.