search

MASILab / 3DUX-Net

244 stars 34 forks source link

How accurate is the direct training on amos2022? #15

Closed Szc-coder closed 1 year ago

Szc-coder commented 1 year ago

Hello, thank you very much for your open source work. I want to ask how much precision can be achieved by training directly on amos2022 task1 instead of using fine tuning? Thank you!

leeh43 commented 1 year ago

Hi, thank you for your interest! We have the training from scratch result in our upcoming paper in a few days. Will send you the link after we put to the arXiv.

leeh43 commented 1 year ago

Our latest paper with the training from scratch scenario in AMOS 2022 is in https://arxiv.org/abs/2303.05785. Feel free to take a look and discuss with me if you have further comments or inquires.

Szc-coder commented 1 year ago

Thank you very much for your work. I have another question about how do you preprocess the amos dataset? Thank you.

leeh43 commented 1 year ago

Please refer to the train_transform of AMOS dataset in the load_datasets_transforms.py. That is the preprocessing of how we prepare the input patches.

Szc-coder commented 1 year ago

Thank you for your patience.