search

liamheng / Annotation-free-Fundus-Image-Enhancement

Project for annotation-free image enhancement of fundus, cataract data.
30 stars 5 forks source link

demo #4

Closed Huangqqqhhh closed 5 months ago

Huangqqqhhh commented 1 year ago

source域和target域是一起训练的嘛，如果是，那么我如何去修复一张未经过放入targret域中训练的白内障眼底图像照片？

Railies commented 1 year ago

你好，我也在关注这个项目，可以交流一下吗

Huangqqqhhh commented 1 year ago

可以的

 

权是海 @.***

 

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年11月1日(星期三) 上午9:47 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [liamheng/Annotation-free-Fundus-Image-Enhancement] demo (Issue #4)

你好，我也在关注这个项目，可以交流一下吗

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

Huangqqqhhh commented 1 year ago

你的联系方式是什么 

权是海 @.***

 

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年11月1日(星期三) 上午9:47 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [liamheng/Annotation-free-Fundus-Image-Enhancement] demo (Issue #4)

你好，我也在关注这个项目，可以交流一下吗

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

Railies commented 1 year ago

抱歉，看不到您的联系方式。我的邮箱:2937898618@qq.com

Huangqqqhhh commented 1 year ago

@.***  

权是海 @.***

 

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年11月6日(星期一) 上午10:36 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [liamheng/Annotation-free-Fundus-Image-Enhancement] demo (Issue #4)

抱歉，看不到您的联系方式

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

Railies commented 1 year ago

额，你的邮箱被*打码了

Huangqqqhhh commented 1 year ago

我加你QQ了

 

权是海 @.***

 

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年11月6日(星期一) 中午11:13 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [liamheng/Annotation-free-Fundus-Image-Enhancement] demo (Issue #4)

额，你的邮箱被*打码了

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

HeverLaw commented 5 months ago

It depends on what model you are using. ArcNet is trained with source data with ground truth and target data without ground truth, so you need to fine-tune or retrain a new model using the source and target data, note that you should add some of the real low-quality images into the target dataset to get a desirable result. However, in SCR-Net and GFE-Net, you don't need to retrain the model, since it can generalize well to unseen images with the desired preprocess.