Among them, 119 are Mikaela's personal privacy pictures.
If you're a very cautious person, maybe you can check your icloud regularly for sensitive photos.
Once hackers intrude into personal cloud storage accounts, a large number of privacy information will be leaked, such as taking private photos of sensitive parts, taking photos of important certificates, recording personal sensitive data files, etc.
So we use weak supervised learning to label training data quickly, and solve the problem of noise and classification conflict between expert rules at the same time.
But collecting enough labeled training data is a very challenging task.
Our code will be open source later.