How can I efficiently label data with holes for YOLO segmentation models?

Hello!

I am a graduate student studying in South Korea.

I am trying to create a dataset using Roboflow to train a segmentation model with YOLOv8.

I have a question regarding the annotation process while building the dataset.

As shown in the attached image, I would like to segment the largest circle (orange layer), but exclude the 5 smaller circles (purple layer).

In other words, during the binary mask creation process, I want to assign 1 to the layer corresponding to the orange circle, and 0 to the layers corresponding to the purple circles.

  1. Is there any preprocessing step I need to follow during the annotation process to achieve this?
  2. If not, should I manually create the mask through post-processing of the segmentation results?

Additionally, do you have any advice for annotating data when I only want to segment circles, as shown in the attached image?

Thank you :slight_smile:

  • Project Type: Segmentaion chargping port
  • Operating System & Browser: YOLOv8
  • Project Universe Link or Workspace/Project ID: