results are different than the paper

[mehrnia]

Hi,
Thank you for providing such a nice instructions to fine-tune sam2 and get medsam2.
I followed the instructions in your Github carefully and hopefully got the result. I did not change any number and I stuck to your instructions.
and trained the model based on your instructions.
what is the reason that the evaluation results that I get is different than the ones in your paper: "Segment Anything in Medical Images and Videos: Benchmark and Deployment"
I attached my results and your results that I'm comparing with:

[JunMa11]

Hi @mehrnia ,

We have run the fine-tuning process twice and the results were similar. Could you please provide more details on your training protocol?

[mehrnia]

Hi,
thank you for the response.
I was wondering what do you mean by the protocol?
I'm follwoing the the instructions below:
https://github.com/bowang-lab/MedSAM/tree/MedSAM2?tab=readme-ov-file#fine-tune-sam2-on-the-abdomen-ct-dataset