I am supposed to start working on a project involving detecting forgeries on pdfs representing document scans (payslips, receipts,…), these documents were forged before being scanned. To be honest, I am a bit lost regarding the approach I should use as I am a beginner on computer vision.
- I suppose I can’t really use methods like the one presented in this repo https://github.com/qcf-568/DocTamper, because it relies on the image being a jpeg that has been forged digitally (I mean, my dataset also contain forgeries made digitally but as it has been scanned meanwhile…).
- My other problem, and it’s not a small one is that the documents I have received are not labelized so I would need a public dataset so I can test some approaches.
Do you have any ideas regarding a method I could explore for my usecase and if possible a dataset to try this on ?
You need to sign in to view this answers
Leave feedback about this