Closed minshu-kim closed 2 years ago
What overlap rate did you adopt for making synthetic datasets used for training when you reported the results to the paper?
I found this in the paper. Warped-COCO datasets are in the range of overlap region 75% and Stitched-COCO in 50%. This work applied Stitched-COCO when training. I solve this issue and close it.
What overlap rate did you adopt for making synthetic datasets used for training when you reported the results to the paper?