I have an existing training dataset with bounding box labels. I converted them to centroid approach and output the required json bounding_boxes.labels format.
The labels are imported successfully but the position in the image looks wrong.
I verified the original label coordinates and also the conversion to centroid and the data output looks just fine.
What are your expiriences with importing existing labels? Smooth sailing or is someone experiencing similar problems?
I can provide more information, screenshots, verification results, … if required