This project concerns the verification of identity by handwritten signatures called "offline" using deep learning methods. In this context, the handwritten signatures affixed to documents were scanned giving rise to grayscale signature images (offline signatures).
This project will first require a state of the art on deep learning methods applied to offline handwritten signatures. It will be a question of understanding and making a theoretical presentation of the methods of the literature.
Then, to apply one of them on databases containing images of handwritten signatures of several individuals. Tests and performance measurements will be carried out to assess the relevance of each method.
This github contains two notebooks which allows us to run our model by training it and testing it on different databases.
There is also a folder named delivrables where you can find several documentation on our project like a detailed report describing thoroughly our work.
The articles forlder is where you can find the multiple related works that we studied in order to make this work.
This work has been inspired by hlamba28