How can I use the official caffe to train your model with the lab color space input

Thanks for your nice work! @richzhang 
How can I use the official caffe to train your model with the lab color space input ? How to deal with the input in the provided prototxt(https://github.com/richzhang/splitbrainauto/blob/master/models/deploy_lab.prototxt)? Using the same RGB frames layer, such as ImageData Layer?