Instance segmentation

Hi, I am trying to train a instance segmentation detectron2 model for a custom data set.
The accuracy of boundary of a mask that I am getting for trained model is low.
How can I improve the accuracy of a boundaries of a mask.
Or is there any other model for segmentation which will give better result for small data set also.

Hi @Rohini_vaidya

For each variation a minimum of 20 samples are needed

For any ML you need to train enough samples else the results or the accuracy might not be as expected