What is it about?
Solving the Online signature recognition problem using convolutional residual networks as a feature extractor and K-nearest neighbor with cosine distance as a classifier.
Featured Image
Photo by Szabo Viktor on Unsplash
Why is it important?
To allow the human handwritten signature to be used automatically in identification and authentication without the need of human intervention.
Perspectives
This publication introduces the best hybrid machine and deep learning architecture for online handwritten signature identification which gave the highest recognition accuracy over all of the state of the art techniques when applied on some famous datasets with the minimum number of training samples.
Professor Gibrael Abo Samra
King Abdulaziz University
Read the Original
This page is a summary of: Using Residual Networks and Cosine Distance-Based K-NN Algorithm to Recognize On-Line Signatures, IEEE Access, January 2021, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/access.2021.3071479.
You can read the full text:
Resources
svc 2004 TASK1 and TASK2 signature verification dataset
This resource contains two datasets TASK1 and TASK2 Each database has 40 sets of signature data. Each set contains 20 genuine signatures from one signature contributor and 20 skilled forgeries from five other contributors. TASK1 contains x, y, time stamp and button status time series data and TASK2 contains three extra time series concerning Azimuth, altitude and pressure data.
FCIT Signature dataset (FCITseg dataset)
Online signature dataset for identification only.
Contributors
The following have contributed to this page
