This topic actually contains two questions which are related, but for the second I did not want to create another topic.
I am looking for a method to detect multiple objects in 1 image also called multi-label classification. I found this topic here on fast.ai but if I understand correctly I can’t do this with a category per directory.
Another thing I have looked at is YOLO which is real-time and thus maybe overkill in my case since I only need to recognize from a photo.
What is my best bet here?
My second question is regarding the classification. I am trying to classify characters which look like text but are not text. The characters will be printed on paper so they will be photographed with different angles and lightning. What is the best solution to regocnize characters? Should I use a CNN or something else like K-NN (I don’t have many classes). And is it even possible to do multi-label classification with something other than a CNN?
Thanks a lot for your time.