search

arcadelab / FastSAM3D

Code for "FastSAM3D: An Efficient Segment Anything Model for 3D Volumetric Medical Images"
https://arxiv.org/abs/2403.09827
Apache License 2.0
68 stars 4 forks source link

Training data set #2

Closed jianjun0407 closed 6 months ago

jianjun0407 commented 7 months ago

Thank you for sharing. Could you introduce the training data set used in FastSAM3D model training?

skill-diver commented 7 months ago

We have list it in the expirement part of the paper, pls check, thank you.

jianjun0407 commented 7 months ago

Indeed, I saw the data set introduction about validation in the paper, but I did not see the data set introduction about train. I will review the paper again later, thank you for your reply

skill-diver commented 7 months ago

We mainly use Brats2021, AMOS, and Totalsegmentator to train.

joseph-bae commented 5 months ago

For BRATS, which modalities do you use? Do you combine all modalities into a single image (4 channels?) Or do you instead treat each modality as a separate sample?

skill-diver commented 5 months ago

We combine them.