How to feed data with a distribution (pdf) as the label?

[Saeed Izadi]

I want to prepare my dataset in a way that each image has a discrete distribution as the label(say 1xK dims with float in [0,1]). How should I configure my arch and data preparation?

[Saeed Izadi]

Yep, I'm already aware of HDF5 data layer, but I remember some issues with it, e.g. 2G size limitation for .hdf5 files. I'm looking forward for a more clean solution. 

On Sat, Oct 15, 2016 at 10:29 PM Seyed Majid Azimi <s9se...@stud.uni-saarland.de> wrote:

as far as I know, using HDF5 you can define many sorts of data as label. Have you already given it a try?

2016-10-15 20:34 GMT+02:00 Saeed Izadi <saeed.i...@gmail.com>:

I want to prepare my dataset in a way that each image has a discrete distribution as the label(say 1xK dims with float in [0,1]). How should I configure my arch and data preparation?

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users...@googlegroups.com.

To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/f2f494d4-e30f-4423-8795-dc9e8bad8a34%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Mengyuan Yan]

Write your python data layer if speed is not your major concern.

[Przemek D]

I think HDF5 is the way to do it. Version 5, according to the developers, can handle files larger than 2GB (version 4 couldn't do this as well as many other things).