About

All the available models are pytorch models trained on free sounds off the internet. Any files you upload will be deleted immediately after classification.

It only uses the first 5 seconds of a sound file to try to classify it, so keep that in mind if you are getting unexpected results.

The Python server is hosted on Heroku, and actually I used a Nodejs server that runs the Python file. Kind of a dumb way to do it, but I already know node and I don't know Flask or Django, so that's what I did.

You can make your own requests to the API if you want. Basically, make a POST request with a file encoded as multipart/form-data with a key of 'sound', and it should return a JSON response with 'type' as the only property. The type is 1 for a happy sound, or a 2 for a sad sound. Optionally, you can specify a model to use. If you make a GET request to the server, it will return a JSON object with a array of the available models under 'models'. The server address is https://happy-sad-sound-classification.herokuapp.com/. Please don't spam it! It usually takes 6-10 seconds to return a response (very slow, I know).