Multi-label classification in caffe

[Belhassen Bayar]

I am a new caffe user and I am trying to implement the multi-label classification in caffe.

I already know how to create a customized data in python like this example but I couldn't find any complete and clear tutorial of how to define the labels when an image belongs to 2 or more classes.

I would be thankful if someone could give me a complete example of how to create such data and how to define the accuracy and loss layers in the prototxt files.

Thx!!

[Bharat Bhusan Sau]

I use hdf5 format for multilabel data. I was finding it more difficult to work with lmdb format for multilabel data.

[Belhassen Bayar]

Could you please explain how do you define the labels for the hdf5 format and most importantly could you please provide the Accuracy as well as the loss layer in the prototxt files.

Many thanks!