yolov8-obb exports to coreml

[Wwshiliu]

Hello，
can the yolov8-obb model be exported as a CoreML model for use in IOS? I tried to export the model, but only when the yolov8n model uses the parameter 'nms=True' when exporting coreml, can it be used normally in IOS. Without the parameter 'nms=True', or 'nms' defaults to False, the model cannot be used for detection in IOS. Why is this? Also, when exporting the yolov8n-obb model with the parameter 'nms=True', the exported model still cannot be used properly. Is this because the yolov8n-obb model does not support the parameter 'nms' when exporting? How should I use the yolov8n-obb model in IOS?

[glenn-jocher]

@Wwshiliu hello,

Currently, the yolov8-obb model may have limitations with CoreML export, particularly with the nms parameter. To ensure compatibility, please verify if the issue persists with the latest version of the ultralytics package. If the problem continues, it might be due to specific constraints of the yolov8-obb model. For further assistance, please refer to our CoreML export documentation.

[john-rocky]

@Wwshiliu
Thank you for your feedback.
Currently, the NMS option only supports Detection models, and in order to use Obb models on iOS, you will need to write the NMS and post-processing in native code.