search

dali92002 / DE-GAN

Document Image Enhancement with GANs - TPAMI journal
GNU General Public License v3.0
181 stars 32 forks source link

Not able to Replicate the stated results (binarization, gray clean-up) on DIBCO dataset and noisy office dataset. #4

Open soumyadeepdey opened 3 years ago

soumyadeepdey commented 3 years ago

The model is not converging for the stated datasets till 300 epochs with the training setup mentioned in the repo. Can you point us to the right settings for training the model?

Ashwani-Dangwal commented 1 year ago

Hi, @soumyadeepdey were you able to train the model successfully using the code provided here?

soumyadeepdey commented 1 year ago

No, I was not able to train the model with this code !!!

Ashwani-Dangwal commented 1 year ago

@soumyadeepdey Thanks for the reply. I also implemented the paper you wrote for color rubber stamp removal and it works great but still there are some edge cases that fails to do the task so did you find any model that works for removing rubber stamps?

soumyadeepdey commented 1 year ago

Thanks for reaching out to me, I have not worked with rubber stamp removal for a long time. Not aware of any recent development/ models in this direction.

On Wed, Jul 12, 2023 at 12:53 PM Ashwani-Dangwal @.***> wrote:

@soumyadeepdey https://github.com/soumyadeepdey Thanks for the reply. I also implemented the paper you wrote for color rubber stamp removal and it works great but still there are some edge cases that fails to do the task so did you find any model that works for removing rubber stamps?

— Reply to this email directly, view it on GitHub https://github.com/dali92002/DE-GAN/issues/4#issuecomment-1631985809, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADDWKP75E6JP6N3ST5IM4QLXPZGH7ANCNFSM4YQQIMHQ . You are receiving this because you were mentioned.Message ID: @.***>

--

Thanks and Regards Soumyadeep Dey mob no : +919433715948

Ashwani-Dangwal commented 1 year ago

Okay. But if you have any insights or direction that can be helpful for me then i am all ears. Thankyou!