Hi,

The test set of your benchmark, all_task_test.txt has 10246 images in it, however, only 9119 of these images are unique, and 1127 are repetitions. Here are some image ids that appear more than once:

000000419974
000000329219
000000279278
000000540502
000000127182

I know that there are many following works that evaluate on your benchmark, so I believe everyone evaluates on this 10246 set.
Has this issue been addressed before? Is there a detail that I am missing?

Thanks in advance.