Hi everyone,
I exported a zipper dataset from Roboflow using polygon annotations and uploaded it directly to platform.ultralytics.com.
Interestingly, I can successfully train Instance Segmentation and Classification models on this dataset. However, when I try to train an Object Detection model, i don’t find any object detection model like yolo26s.
Since polygon data inherently contains bounding box boundaries, shouldn’t the platform auto-convert polygons to bounding boxes for object detection? Why is object detection failing while segmentation works seamlessly?
Any insights or workarounds would be highly appreciated!