Train a classifier to detect objects based on its form and colors
Hello Everyone!
I'm trying to create a birds classifier, based on shape and colors of the birds. In the examples that I found, the haar classifier does not consider the colors in the training, only the shape. My question is: How can I create a haar classifier that detect birds based on shape and colors of the birds? It is important to say that the classifier must distinguish the different types of birds, based on its shapes and colors. There are kinds of birds that have the same or very similar form, but the colors are different, and based on color difference that the classifier must recognize what kind of bird it is.
Thank you! Renan Rosa
You need the ICF/ACF detector of Piotr Dollar. It is basically cascade classifier training with more channels, including color. Are you limited to OpenCV (which doesnt have a good implementation) or can you use other frameworks also? Limited to C++?
@StevenPuttemans in fact, I think the ICF/ACF detectors are completely removed from the contrib branch. This PR seems to remove everyhing related to the channels computation :S
Hmmm that is weird :D Didnt know about that .. however, there are publicly available C++ implementations like this. But to bad they deleted it ...
Yep, I haven't understood yet why it has been removed, and without any clear note about it (even the docs still refer to the ICF/ACF... I had to navigate through the PRs to find out what happened...). Btw, here there's also a C++ GPU implementation.
Thanks for the answer guys!
I will try the implementation that LorenaGdL said.