I am working on detection of fraud on cheque images, I want to detect if there is fraud in handwritten cheque basically I want to predict fraud if there is more than one handwriting style present and non-fraud if there is one handwriting style present in amount,payee.
How should I approach this problem and any datasets available for this problem other than cval database?
I dont know about datasets but maybe using a pretrained model and removing some layers and then perform further training on cheques, maybe you can get some good performance! Maybe YOLO or someother model might be good for this…