Training haar classifier to detect each book on a bookshelf [closed]
Hello, and thanks in advance for taking the time to read my first post on the forum. I started playing around with Node openCv two weeks ago and started to build a book recognition script. So far I came to the conclusion that I need to train my own classifier to reach high success rate on image recognition and I'm now sticking to this tutorial to achieve it: http://coding-robin.de/2013/07/22/tra...
My only question so far is that, considering the following image:
how shall I process to train the haar classifier? Inputing the whole image or obstructing every book but one so the classifier can 'learn' what a single book is? Most generally, is there a way to train the classifier how to detect multiple models(occurences of the same king of object) on a single image or shall I process object by object (book by book).
Thanks in advance for your answers, have a beautiful day.
if you want to find out, which book it is, the cascade classifier is the wrong tool, then.
(also, that blogpost is horribly outdated, and the node wrappers are not at all supported from opencv)
@berak thanks for your answer. Actually, I don't need to know which book it is. All I want to come up with is getting coordinates of each book on the image with high success rate. What would be the best option to make this happen in your opinion?