OpenGVLab / SAM-Med2D

Official implementation of SAM-Med2D

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Partitioned datasets

Rey-nard opened this issue · comments

Hi, I noticed that the SAM-MED2D-20M dataset wasn't partitioned into training and testing sets. I would like to evaluate the model performance on a testing set. I wonder if it's convenient to provide the information about the testing set?

During testing, we randomly divided all datasets into 20% for testing based on the image index. Since not all of our currently open datasets are available (only those permitted are open), we suggest that you also adopt our method of randomly dividing the dataset for training and testing.

Thanks for your explanation : )