Dataset enquiry

Open petertran1811 opened this issue 3 years ago • 3 comments

Thank you for your excellent work. Could you provide the information about the 120000 dataset used in the paper? I randomly take images from the set, however, the performance is not so good.

Dec 08 '22 04:12 petertran1811

Hi, I don't have the accurate file list. How did you randomly take images? Please try to follow the experiment instructions about CBN and use all sub datasets.

Dec 08 '22 05:12 tyz1994

Thank you for your reply. I randomly took 500 ids from the unrrealv1_2 and trained them on Vision transformer backbone. I have done the same thing for 500 ids taken from market and msmt dataset. By evaluating in the same test set, the model trained on unrealv1_2 was far poorer than the other twos.

Dec 08 '22 11:12 petertran1811

unrealv1.2 has only one scene and one person type. It's not a surprise that only using one subset doesn't show the advantages of unreal synthesis.

Dec 08 '22 11:12 tyz1994