Handwritten Text Analysis and Identification using Deep Learning Models

Handwritten text analysis for author identification is a significant but difficult topic with wide applications across several areas. The effectiveness of deep learning models, notably AlexNet and ResNet-101, in identifying the writers of written samples is investigated in this work. This study has experimented with these algorithms to obtain discriminative features for purposes of identification, focusing on text images from various persons. As a part of the research process, the models were trained on a variety of handwritten sample datasets and their performance was assessed using relevant metrics. Using a variety of handwritten text datasets, this study explores writer identification using deep learning models, namely AlexNet with 95.46% accuracy and ResNet-101 with 96.02% accuracy. The models’ promise in forensic analysis and historical document authenticity is highlighted by the use of varied handwritten samples in training and evaluation.