search

ljwztc / CLIP-Driven-Universal-Model

[ICCV 2023] CLIP-Driven Universal Model; Rank first in MSD Competition.
Other
521 stars 58 forks source link

For the results of BTCV in the reproduction paper #44

Closed Adoreeeeee closed 7 months ago

Adoreeeeee commented 8 months ago

For the results of BTCV in the reproduction paper, I have some questions I want to ask, the 5-fold cross-verification method you mentioned, is it done five times with train.py and val.py? According to the classification in BTCV.json. Looking forward to your reply

ljwztc commented 8 months ago

Yes. We trained five different models according to the split in BTCV.json, but the code is not exactly same with train.py and val.py.

AustinYuAo commented 7 months ago

Hi, sorry to bother you. I want to reproduce the Swin UNETR model, how many epochs do I need to train, is the preprocessing the same as https://github.com/ljwztc/CLIP-Driven-Universal-Model/blob/2239db89c61daa985f4c6c11a10fd3e7f7b0d6e5/dataset/dataloader.py#L225C22-L225C22? Do I need to be aware of anything else?

ljwztc commented 7 months ago

We conduct five-cross validation according to this file. More details please refer to https://github.com/ljwztc/CLIP-Driven-Universal-Model/issues/26

AustinYuAo commented 7 months ago

Thank you very much for your response. We have also submitted our training results to the official website, but we haven't received any results yet. Can I ask if you know how to obtain the segmentation results? I am eagerly looking forward to your reply.