Hi @khatami.mehrdad ,
I trained a custom YOLOv8 model and achieved a confidence score of 0.9 during inference. I then converted this model to a DeGirum-compatible format using DeGirum AI Hub. However, after conversion, the model now only yields a confidence score of 0.6.
Is there a way to resolve or improve this confidence drop after the conversion?
Hi @manasa
Thanks for reaching out.
Could you please provide more details, such as the JSON file used in your PySDK model configuration? If possible, also share the .pt
file of your trained model along with the validation dataset. This will allow us to reproduce the issue on our side and better investigate the confidence drop after conversion.
Looking forward to your input!