pickle VideoLikeContainers when data dtype is not supported by hdf5 format

fixes #18 (closed)

added 1 commit

• d9e65a6f - pickle VideoLikeContainers when data dtype is not supported by hdf5 format

Compare with previous version

enabled an automatic merge when the pipeline for d9e65a6f succeeds

merged

mentioned in commit 8cace2f8

@amohammadi is there a better solution for this ?. If not, I will patch the pytorch dataloader with this as well.

I could not find a better solution, and this change is backwards compatible. You should always load these files with VideoLikeContainers.load() even if you use them in a pytorch dataset.

ok. What is the best way to load a single image from a videoLikecontainer, __getitem__ method ?

yes, VideoLikeContainers.load()[index]

If the saved file is an hdf5 file, it will not load the whole thing into memory.

also call np.asarray() on output to be sure.

mentioned in issue bob#269 (closed)