[peterhj]

Bonus round:

(1) Find the ImageNet (ILSVRC2012) images possessing EXIF orientation metadata.

(2) Of those images, find which ones have the "correct" EXIF orientation.

[yeldarb]

Has anyone done this yet? I'd be curious to know if their dataset is "cleaned" or not.

[peterhj]

Spoiler, ish:

The last time I measured ImageNet JPEGs with EXIF orientation metadata, the number of affected images was actually quite small (< 100, out of a dataset of 1.28M). There are also some duplicates, but altogether it seems fairly "clean."