Open EcxaByte opened 1 month ago
Hi, thank you for your interest, and sorry for the delayed response.
I ran experiments with centerlines using MapTRv2. One can see similar behavior as for the other classes.
However, I can not guarantee whether the centerline results I have are exactly correct in terms of using the latest version of the splits etc, so if you need exact mAP numbers I would recommend re-running those experiments on your own.
Thanks for the great work! I just wanted to know if you have some results for the mAP for the new geographic splits with regards to the centerline for the different models that are not mentioned in the paper? Thanks in advance!