Contamination detection for fruit classification with PyQt UI.

So I need a detection system to detect contaminants in the class. For example, I have a basket of apples, but there is an orange in the basket, so I will label the orange as a contaminant.
Requirements:
1. training data for each class maximum 100 images.
2. outomatic re-train if new class found (I will explain later)
3. use LBH feature extractor method, albumination or other augmentation method to change the contrast and brightness of the image.
4. After getting the features from step 3, the classification process is carried out using random forest.
5. code in python
If you are interested please contact me.