I am using caffe deep network library and I have to use googlenet in my data (images).
The problem is the fact that I want to do a lot of pre-processing operations, and I don't want to save every image processing operation as image files and then execute create_imagenet.sh included in the library. This library needs text files to indicate where are the train and validation image files.
I don't want to save a lot of images and then converting them in LMDB.
All I want is to do the image processing operation in a series of images and save them in a row in a lmdb file to be read by caffe later.
Is that possible?
This github repository might be the one you are actually looking for: https://github.com/kyamagu/matlab-lmdb