Using Caffe for Videos. Trouble with getting data into caffe

[mnik...@gmail.com]

Hi, 

I am trying to use caffe for learning features for classification in videos by increasing number of channels. While i have my data in lmdb format, i face a problem as the number of frames per video is not the same. I would like to know when a video ends as during prefetch, I wouldnt want it to append frames of the previous video to the next video. 

Also, I would like to know if this could cause problems, as i not too sure on how lmdb works. I am aware that it has a key associated with it, however videos belonging to the same class would have the same key. 

Or, if using HDF5 would be better, if so, how should the data be made available to MATLAB for frames extracted from videos.

Any help would be appreciated 

Thanks

[Paolo]

Having the same problem with monodimensional data. I cannot find any solution...