The smark speaker demo shows Noise Reduction, Audio Fingerprinting, Wake word detection, Vad detection, and Text Classification
To run the fingerprint part of the demo, run the docker-compose up within the dejavu_repo. Then add some mp3 files within the dejavu_repo/mp3. Then use the docker_id with it. The one shown during the demo was "Field of Hopes and Dreams Remix". A good way to test quickly is via youtube-dl
To run the general classifier, check out the README.md within general_classifier and download the required model files
There are 2 notebooks that you can check out. the smart_speaker demo and the classifier notebook.
For link to the slides, https://docs.google.com/presentation/d/1X3Brj00kGBwNq0OSHT47DVG3nRY0FQzISy2TIan0GrI/edit?usp=sharing