Size, shape and small differences in classification ...

[MyVanitar]

Hello,

Consider we have some apples and we want to classify them. apples could differ in shape, size or sign of bruise on them.

Does deep learning classification could be able to distinguish these if we provide the related images for each class?, I mean something like this:

class-1: big apple
class-2: small apple
class-3: bad apple
class-4: good apple

Besides, consider we have an image which consists of three apples. Left, middle and right. how can I make the classifier to provide the classification results for these three apples as for example: left: good apple , middle: bad apple , right: big apple

[humphd]

In my experience, yes, you can train on any set of categories. The categories can be made up by you ("good apple" vs. "bad apple"), you just need enough samples of your category in order to train it to recognize another one that's similar. I've been working on a system that decides whether an image is "interesting" or "not-interesting" according to a set of criteria that are important to my data and app, and it works amazingly well. These two categories are obviously made up, and only exist as a set of training data that I've assembled.

[MyVanitar]

1. The thing is in my case, the difference is small, for example big apple or small apple. also a bad apple could just have a small bruise on it. Does this work for such cases also?

2. besides I want to know if the classification can handle the quantity of the desired object in the image. for example if it contains two apple, it should mention both with its confidence percentage. Somebody told me classification does not do this, you have to use segmentation. is that true?

[Gowon420]

The issue of Classification is still giving me problem because i am still a beginner in Machine Learning

What do you think i should do?
What book can you recommend for my fast understanding??