is this model can be used to do image deburing ?
Opened this issue · 4 comments
sunzhe09 commented
is this model can be used to do image deburing ?
zzr-idam commented
没问题，效果也是很好的。
…----------
该邮件从移动设备发送
--------------原始邮件--------------
发件人："sun ***@***.***>;
发送时间：2024年2月18日(星期天) 上午9:59
收件人："zzr-idam/UVM-Net" ***@***.***>;
抄送："Subscribed ***@***.***>;
主题：[zzr-idam/UVM-Net] is this model can be used to do image deburing ? (Issue #1)
-----------------------------------
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>
sunzhe09 commented
HI,how to test with my custom blurry image?
zzr-idam commented
We provide the training code, first this model should be trained on your customized dataset and then tested.
zzr-idam commented
There is no need to reduce the number of layers in the UNet, my basic module has three SSMs, you can remove the one that is the most computationally intensive (after reshaping); besides, you can conduct a ReLU or Sigmoid on the output of the UNet and do a residual or multiplication operation with the original input image.
output = nn.ReLU()(UNet(x))
output = output * x - output + 1 // AODNet equation
return output