Training Multi-labelled Image dataset

BVLC / caffe

Caffe: a fast open framework for deep learning.

http://caffe.berkeleyvision.org/

Other

34.04k stars 18.7k forks source link

Training Multi-labelled Image dataset #656

Closed ankurhanda closed 10 years ago

ankurhanda commented 10 years ago

Dear all,

I would like to use conv-nets on NYU dataset http://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html where instead of one label per image you have multiple labels in an image i.e. per pixel labels for each image in the training dataset. Can somebody point me to any example in caffe that does that? How should one set up the first data layer? Is top: images_label exactly for this kind of stuff I'm looking for?

Regards, Ankur.

ssierral commented 10 years ago

I think this has been solved in the dev branch and you can have some information here https://github.com/BVLC/caffe/issues/460 and here https://github.com/BVLC/caffe/pull/523

shelhamer commented 10 years ago

Closing -- this will be addressed by #523.