Hi. Thanks for the great paper and repo.
One phenomenon I see when running your script in the README.md is that the HOTA score of DPM, FRCNN and SDP are different, even though they should be the same. Their ground-truths gt.txt files also look the same.
Can you have a look?
Hi. Thanks for the great paper and repo. One phenomenon I see when running your script in the README.md is that the HOTA score of DPM, FRCNN and SDP are different, even though they should be the same. Their ground-truths gt.txt files also look the same. Can you have a look?
Command
Output