Feature request: Seed value concept

Question

ojwenzel opened this issue 5 years ago · 2 comments

For usage in science, reproducibility of the corruption types would be very helpful. Thus, some sort of seed value concept would be a great feature.

Answer 1 · 2019-11-14T14:29:18.000Z

Resetting the numpy seed before calling the corruption function does yield reproducible results for all corruption types :)

Answer 2 · 2019-11-18T16:00:07.000Z

For clarification (if somebody else wants to use that as a workaround for the time being):

np.random.seed(131213)
corrupt(img, corruption_name=corruption_name, severity=severity)

yields a corrupted image that is reproducible (with the same seed)