search

MrGiovanni / AbdomenAtlas

[NeurIPS 2023] AbdomenAtlas 1.0 (5,195 CT volumes plus nine classes)
https://www.cs.jhu.edu/~alanlab/Pubs23/qu2023abdomenatlas.pdf
MIT License
182 stars 13 forks source link

About nnUnet pre-trained model #3

Closed Betty-J closed 8 months ago

Betty-J commented 9 months ago

Hi, Thanks for sharing! I noticed that you provided two pre-trained models for download. So regarding nnUnet, will pre-training models be provided? I saw in your article that you used three models including nnUnet.

Chongyu1117 commented 9 months ago

In the "Implementation Details" section of Appendix A, it is mentioned that nnU-Net is trained on 47 CT volumes from the BTCV dataset. For those interested in replicating the training process, the BTCV dataset can be downloaded, and the weights can be trained independently. Alternatively, the nnU-Net GitHub repository offers the convenience of downloading pre-trained weights specifically for the BTCV dataset. Please be aware that the pretrained weights from nnU-Net v1 are not compatible with the latest version, V2. Unfortunately, pretrained weights options are not yet available for V2. I recommend following the structure with nnU-Net v1 you can run 

nnUNet_download_pretrained_model Task017_AbdominalOrganSegmentation

To download the pretained weights on BTCV