torchrs icon indicating copy to clipboard operation
torchrs copied to clipboard

fair1m Only small part 1 dataset

Open robmarkcole opened this issue 4 years ago • 3 comments

I noticed the official dataset is split into parts 1 & 2, with the bulk of the images being in part 2

image

The data downloaded using the script in this repo only downloads a subset of 1733 images, which I believe are the part 1 images?

robmarkcole avatar Sep 15 '21 11:09 robmarkcole

That's correct. I haven't updated it in a month or so, so they may have released more imagery.

isaaccorley avatar Sep 15 '21 13:09 isaaccorley

The part2 images are 42.94 GB on disk for 14,757 items. I guess 1 million images will be a gigantic dataset on disk

robmarkcole avatar Sep 15 '21 16:09 robmarkcole

How do I download the full information at this time?

jsxyhelu avatar Jan 31 '23 11:01 jsxyhelu