Are the greater than 50 scores from a single model?
Update: Yes atleast slightly above the 50 score , don't know about the 51s and 54s , but I will update you when I get there hopefully
most likely ensembles of strong models. single model scores should probably be in the 0.48 to 0.5 range.
still searching for a good model...
ensembles though are not going to win this competition due to the t4 and time restrictions, so if the scores are from a single model then that's so impressive, hopefully they are going to respond
are You using a YOLO model? am using fastercnn from torchvision and still can get pass 0.18
Switch to yolo then. Others are also there like detectron2.
I also don't know how they are reaching the 50 score. It's impressive.
great
most likely ensembles of strong models. single model scores should probably be in the 0.48 to 0.5 range.
still searching for a good model...
ensembles though are not going to win this competition due to the t4 and time restrictions, so if the scores are from a single model then that's so impressive, hopefully they are going to respond
are You using a YOLO model? am using fastercnn from torchvision and still can get pass 0.18
Switch to yolo then. Others are also there like detectron2.
I also don't know how they are reaching the 50 score. It's impressive.
great