concerning preprocessing the data ? about resnet-18-caffemodel-on-imagenet HOT 5 CLOSED

Hi, shrt 256 was settled before creating LMDB. In another word, raw data were preprocessed to 256xN(Nx256).

from resnet-18-caffemodel-on-imagenet.

It has already been contained in official repo : your-caffe-master/tools/extra/resize_crop_images.py.
To be honest, we have not implemented other data preprocessings, yet.
It takes about 326G and 13G for training and validation dataset, respectively.

from resnet-18-caffemodel-on-imagenet.

Thanks alot, can you also share the script by any chance?
by the way, did you run any other (fancy) preprocessings in advance? such as those implemented for inception based architectures?
and finally how much space did those lmdb take ?

from resnet-18-caffemodel-on-imagenet.

Thank you very much, I really appreciate it :)

from resnet-18-caffemodel-on-imagenet.

:)

from resnet-18-caffemodel-on-imagenet.