How do I generate Forged Image using Genuine Signature image?

I am writing a code to generate fake signature images that looks almost same to genuine image. My problem is that, I have to train my classifier with my own dataset. But my dataset doesn't have forged images hence, I have to manually generate it.

Until now I have tried one concept that segment an image into different chunks like 10 rows and 10 columns then apply affine transformation and then save that chunk as new image's part. The results are not that good.

Image on right is Genuine and on left is Forged.

Divide rows then apply transformation:

Divide cols then apply transformation:

https://stackoverflow.com/questions/66938495/how-do-i-generate-forged-image-using-genuine-signature-image April 04, 2021 at 02:15PM