search

jeya-maria-jose / UNeXt-pytorch

Official Pytorch Code base for "UNeXt: MLP-based Rapid Medical Image Segmentation Network", MICCAI 2022
https://jeya-maria-jose.github.io/UNext-web/
MIT License
468 stars 78 forks source link

Parameter Settings on BUSI dataset #12

Closed xiaohancl closed 2 years ago

xiaohancl commented 2 years ago

Can you share your parameter Settings on the breast ultrasound image dataset?I set the epoch=400, batch-size=8, lr=0.0001,momentum=0.9,optimzer=adam and scheduler=CosineAnnealingLR and channels setting as [16, 32, 128, 160, 256],but got the result:Dice:75.74 IOU:61.99(In the paper, it's 66.95 ) Thanks！

jeya-maria-jose commented 2 years ago

Hi @xiaohancl , I just rechecked and I am able to get the reported performance in the paper. The settings you used are correct but did you resize the images to 256? Also, for the experiments on BUSI dataset, we only considered the benign and mailgnant images resulting in a total of 647 images.

jeya-maria-jose commented 2 years ago

Hi @guangguangLi , which line are you getting the error?

xiaohancl commented 2 years ago

Hi @xiaohancl , I just rechecked and I am able to get the reported performance in the paper. The settings you used are correct but did you resize the images to 256? Also, for the experiments on BUSI dataset, we only considered the benign and mailgnant images resulting in a total of 647 images.

i have resized the images to 256 by the train_transform and val_transform acording to your source code. Busi dataset is also downloaded following the link you provided. And then put the 647 images in the inputs/busi/images and put the 647 masks in the inputs/busi/masks/0

xiaohancl commented 2 years ago

Hi @xiaohancl , I just rechecked and I am able to get the reported performance in the paper. The settings you used are correct but did you resize the images to 256? Also, for the experiments on BUSI dataset, we only considered the benign and mailgnant images resulting in a total of 647 images.

i have resized the images to 256 by the train_transform and val_transform acording to your source code. Busi dataset is also downloaded following the link you provided. And then put the 647 images in the inputs/busi/images and put the 647 masks in the inputs/busi/masks/0

xiaohancl commented 2 years ago

I have resized the images by train_transform and val_transform(Resize(256,256)).dataset is begin and maligant(total 647)

---Original--- From: "Jeya Maria @.> Date: Thu, Apr 28, 2022 01:04 AM To: @.>; Cc: @.**@.>; Subject: Re: [jeya-maria-jose/UNeXt-pytorch] Parameter Settings on BUSI dataset (Issue #12)

Hi @xiaohancl , I just rechecked and I am able to get the reported performance in the paper. The settings you used are correct but did you resize the images to 256? Also, for the experiments on BUSI dataset, we only considered the benign and mailgnant images resulting in a total of 647 images.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>