@WongKinYiu
Hi，thank you very much for updating the new performance on main branch.
I've tried to training yolor-p6 on coco2017 from scratch with the main branch, but it only can get about map 52.7% on test-dev2017, same as the pretrained model provided on main branch, is much lower than the new performance of yolor-p6 map 55.7%.
So, is there any code update of main branch. I would like to achieve the new performance to go on my project.
Also, would you like to release to yolor_e6.cfg and yolor_d6.cfg of main branch?
Thank you very much for this perfect work.
@WongKinYiu Hi，thank you very much for updating the new performance on main branch. I've tried to training yolor-p6 on coco2017 from scratch with the main branch, but it only can get about map 52.7% on test-dev2017, same as the pretrained model provided on main branch, is much lower than the new performance of yolor-p6 map 55.7%. So, is there any code update of main branch. I would like to achieve the new performance to go on my project. Also, would you like to release to yolor_e6.cfg and yolor_d6.cfg of main branch? Thank you very much for this perfect work.