0

In ClearML Dataviews, it is possible to add augmentations.

Where do these augmentations run?

Options

  1. Original data gets downloaded to local, then runs (on which device? How is multiprocessing handled?)
  2. Only augmented data gets downloaded to local cache, augmentations run remotely (who pays for compute? How fast? Should pipelines be changed accordingly?)

I couldn't find this in the docs.

Gulzar
  • 23,452
  • 27
  • 113
  • 201

1 Answers1

0

Disclaimer: I'm one of the team members of ClearML

Unfortunately since this is done client side, the easiest is to take these instructions (as basically this is what they are), and use imgaug or albumentations to implement real-time augmentation on client side

Martin.B
  • 599
  • 3
  • 9