by Louie Pascual, John Dinnex Valeros, Al Arazas, and Allan Sioson
In this work, we used an artificial neural network to do offline specimen signature recognition and achieved 99.47% accuracy given the 940 raw images of signatures of 79 authors from the SigComp 2009 dataset.
The repository only includes the C++ programs used in extracting the features from the images, generating the datasets, and producing the ANN models. The extracted datasets will be uploaded soon.