2

I currently got a yolov5 dataset , with everything on it (labels in form of : label , x , y , widh , height).

My question is , is there an fast way to convert it into a proper custom dataset for mask-rcnn?

Im currently searching the whole web, otherwise i have to make a custom dataset from scratch with about 1000 pictures and i sadly i got limited time.

Otherwise i would sit down and use this guide: https://engineering.matterport.com/splash-of-color-instance-segmentation-with-mask-r-cnn-and-tensorflow-7c761e238b46

Murphy
  • 21
  • 2
  • from what I understand you have just the bboxes, but you want the segmentation (/ polygon) points? Unfortunatelly for that you will need to make annotations by hand - the conversion can work the other way (segmentation to bboxes) but not the way you need (not enough data in the bboxes...). please share if you figure it out and I am wrong - would be very helpful and save a lot of time in the future. – ArieAI Feb 09 '23 at 15:27

0 Answers0