deblur-gan
deblur-gan copied to clipboard
RaphaelMeudec
Input image resolution.
The network has been trained with this input shape so the input shape of the network is fixed. An option to perform on larger images would be to cut the images into smaller images of 256, 256 and unify the outputs
I have one more issue. How do I select the model to be used? The code performs poorly when there's a moving object in the image. How do you justify that?
I tried doing that! But the problem is that the features of the image is lost. Hence, the results are not good. Here's an example.
On Mon, Jun 11, 2018 at 6:33 PM Raphael Meudec [email protected] wrote:
The network has been trained with this input shape so the input shape of the network is fixed. An option to perform on larger images would be to cut the images into smaller images of 256, 256 and unify the outputs
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/RaphaelMeudec/deblur-gan/issues/16#issuecomment-396235624, or mute the thread https://github.com/notifications/unsubscribe-auth/AWH51nhiWM3aHnGo-QcTHUHd6bdXyp2wks5t7mqLgaJpZM4UiGNt .
I tried doing that! But the problem is that the features of the image is lost. Hence, the results are not good. Here's an example. … On Mon, Jun 11, 2018 at 6:33 PM Raphael Meudec @.***> wrote: The network has been trained with this input shape so the input shape of the network is fixed. An option to perform on larger images would be to cut the images into smaller images of 256, 256 and unify the outputs — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#16 (comment)>, or mute the thread https://github.com/notifications/unsubscribe-auth/AWH51nhiWM3aHnGo-QcTHUHd6bdXyp2wks5t7mqLgaJpZM4UiGNt . Hi shivam I have trained a model on Go pro dataset.But iam getting bad results.can you tell me why? I will show you an example image
