Hi,
I need to do classification on both training and validation sets (totally 150k images) for every snapshot versions of the CaffeNet. I do snapshotting after every 1000 iterations.
So far, I have been doing this with raw images (.jpg format) and it takes a lot of time.
I am wondering whether there can be another way of doing classification using LMDB format in C++ as I think it will be faster
Thank you