• DepthAI-v2
  • How can I run yolov8 instance segmentation at OAK?

I am using the RGB & TinyYolo with spatial data for example (https://docs.oakchina.cn/projects/api/samples/SpatialDetection/spatial_tiny_yolo.html). First, I trained Yolo v8 object detection in my custom dataset. And convert to a blob file using https://tools.luxonis.com/ . That works pretty well!

But, I would like to run a custom instance segmentation model on OAK. The problem is that the conversion tool supports detection only.

Any thoughts about how could I do that?

10 months later

u111s No, I employed a standard YOLO object detection model alongside a secondary neural network for instance segmentation, like the Segment-Anything model or Google's MediaPipe MagicTouch. This approach proved to be effective for my specific problem.