Sound classifier
Visit Sound classifier live demo!
A machine learning (ML) model for classifying sounds as happy or sad. It doesn't work extremely well, but considering that I trained it on my laptop with about 100 sounds, I'm pretty happy with its performance. I used PyTorch with torchaudio for training and running the model. I also made a website with the model hosted on Heroku to make it easier to use. Unfortunately, Heroku no longer offers free dynos, and it no longer works because I don't want to pay for a dyno just for this.
The ML part is written in Python, but when I was setting up the website I decided I didn't want to learn Django or Flask, so I wrote the server in Node.js and simply had it spawn a Python process to run the model. It's a terrible and slow way to do it, but it works. (At least, until Heroku ended their free tier). I'm trying to get it to run on Fly.io, but it seems that Fly.io doesn't support multiple build packs like Heroku, I'll have to rewrite it to be entirely in python or package the python file as an exe or something.
Type: Website
.
