Training Mask RCNN without BBox
ruyadisk opened this issue · comments
Jyun-Wei, Chen commented
Hi there, is it possible to train instance segmentation model based on Mask RCNN with coco format dataset without bbox data? (already got segmentation data)
NICKQZL commented
好像不行,自己写个脚本根据polygon最小x,y和最大x,y算框就行了