Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

Hi @rijobro , I got three suggestions and was not sure about if they are helpful. Maybe we can discuss about them.
1, In this part of the video dataset, I think you transfer a bgr image to rgb one implicitly. Could it be better if a parameter like "color_type", default as "rgb", are given and used to decide whether the image channel should be changed.
https://github.com/rijobro/MONAI/blob/c3fd10aeda4bf488f98ad2850c12978fa85e1f42/monai/data/video_dataset.py#L56
2, In the "video_seg.ipynb" tutorial, would it be helpful if the training dataloader's "shuffle" parameter were set to "True"?
3, The timing part in "video_seg.ipynb" may not work when cuda device involved，since cuda is a asynchronous process. The link below is a reference on how to do it.
https://pytorch.org/TensorRT/_notebooks/EfficientNet-example.html

@binliunls Changes have been incorporated.

Depends on the Video dataset PR in MONAI.

Thanks.