Handwriting Recognition System Using YOLO and CTC
Neeraj Kumar Garg, Neelam Sharma, Gautam Jain, Vipul Jain, Vaibhav Upreti · 2023
In a multitude of applications, handwriting recognition holds significant importance, including tasks like transforming handwritten documents into digital format, automating form processing, and intelligent character recognition. This paper introduces a novel handwriting recognition system that integrates the You Only Look Once (YOLO) object detection algorithm with the Simple Handwritten Text Recognition (HTR) model. The YOLO algorithm is widely recognized for its ability to perform real-time object detection. By adapting it to the task of handwriting recognition, we can detect and localize handwritten text regions within images. The YOLO algorithm is trained on a large dataset of annotated images, enabling it to learn features specific to handwriting. Once the text regions are detected, the HTR model is employed for accurate recognition of the detected text. HTR is a deep learning-based approach that leverages recurrent neural networks (RNNs) to transcribe handwritten text into machine-readable format. The model is trained on a diverse set of labeled handwriting samples, enabling it to effectively recognize and convert handwritten characters into digital text. Our proposed system combines the strengths of YOLO's robust object detection capabilities with HTR's accurate text recognition. The integration of these two components allows for efficient and reliable handwriting recognition, even in challenging scenarios with varying handwriting styles and image qualities. To evaluate the performance of our system, we conducted experiments on benchmark datasets and compared the results with existing approaches. The experimental results demonstrate that our system achieves superior performance in terms of both detection accuracy and recognition accuracy, surpassing state-of-the-art methods. Overall, our proposed handwriting recognition system utilizing YOLO and HTR presents a robust and effective solution for converting handwritten text into digital format. This system can be applied in numerous real-world applications, contributing to the automation and digitization of handwritten documents, streamlining administrative processes, and facilitating efficient information retrieval.