GPU accelerated cascade classifier training, is it in the pipeline? Anyone has tried this before?
I was wondering, if there is an intention of creating a GPU powered pipeline or a decent multicore powered pipeline for object model training using the cascade classification interface. For now my knowledge of the CUDA language is too poor to actually consider implementing it myself, but it could be a future project.
I think that this could push the training of actual object models alot, seeing that recent publications succeed in training decent models in less than a day.