Hello. I have the following problem: I have made a frame of images in CVAT, I used a rectangle for marking (the object is asymmetric, both its ends are not similar to each other), when marking the angle was from 0 to 360 degrees. After that I converted the dataset and started training the YOLOv8-obb model. I have only 1 class in my project and the object is perfectly detected and the rotation angle is perfectly determined, but at points between 180 and 0 degrees an anomaly occurs and completely incorrect angle readings are displayed. I would like to see the angle either from 0 to 360 or -180-0-180. How can I solve this problem? Please help me figure it out.
The range of OBB is 0-90. It can’t predict angles beyond that. Angles beyond that degenrates to a range between 0-90 due to how it works.