The idea is rather simple and feasible, but as of this moment, not really useful. It just popped up in my mind, and I decided to write it down in case someone is looking for a ML challenge to work on. The idea is very simple:
- Use the iTunes Search API to download app screenshots and categorize them by folders, corresponding to the names of categories that those apps have been featured in.
- Teach a simple convolutional neural network (CNN) to categorize apps based on screenshots provided
- Try and evaluate the NNs capabilities using screenshots that are not present in the model.
That's it. Simple, but not really useful.
P.S. Please, drop me a comment, if you are intending to work on this. I'd love to see the results.