I'm trying to think of ways I could use machine learning to enhance my apps. Feedback wanted!
You can find my apps here.
- Natural language processing to parse a new event title, like Fantastical.
- Generate color palettes from images using deep learning: http://colormind.io/blog/
- Generate color palettes based on the user's preference: https://medium.com/demagsign/how-khroma-uses-machine-learning-to-create-endless-color-palettes-c00fff9247ed
- Improve favorites search. Favorites are stored locally and have its own search. Since Giphy doesn't expose keywords in their API, we can only search the GIF titles. We could use machine learning to find out what's actually in the GIF and generate keywords based on that.
- Learn what the user usually searches for and likes and then make a "for you" feed.
- Use machine learning to detect faces and add a menu item to quickly redact faces. (Vision framework)
- Use machine learning to detect text and add a menu item to quickly redact text. (Vision framework)
- Audio noise reduction using machine learning: https://towardsdatascience.com/real-time-noise-suppression-using-deep-learning-38719819e051
- Transcription.
- Optimize the crop based on faces and other main subjects in the photo. (I'm already working on this)
[Black Out] - Add an invisible filter to make it so image-processing neural nets cannot properly process the image.
I remember seeing something where tiny non-noticeable changes could be made to images so that object-classifying neural nets see them as something they are not (i.e. a photo of people classified as a dog).
It would be nice to have photos on the internet publicly without worrying about companies scraping them, identifying you and keeping that data.