Hi there,
I have trained custom models (v11n/v8s) within ultralytics hub and they both work flawlessly when tested in the ultralytics hub IOS app. When using the downloaded CoreML version of the same models (nms=true) in a plain xcode-project with this YOLO dependency using YOLOCamera() it detects anything but the actual objects. Boundingboxes are all over the place. I attempted this, after I ran into the same issue using the model(s) in my own IOS-app, extracting results as ModelnameOutput and when extracting them as VNRecognizedObjectObservations.

Is there a crucial postprocessing step I could've overseen? Does the ultralytics hub app use a different detection extraction approach?