Abstract:Optical Character Recognition (OCR) stands as a pivotal technology in the digitization and processing of textual information from images. In this study, we propose a novel approach to OCR leveraging the VGG19 convolutional neural network (CNN) architecture. VGG19, renowned for its depth and performance in image classification tasks, is repurposed here to tackle the intricate challenges of character recognition. Through extensive experimentation and evaluation, this study demonstrate the efficacy of the proposed approach in achieving state-of-the-art accuracy and robustness in extracting textual information from images. This study uses diverse datasets comprising printed and handwritten text samples, augmenting them using various techniques to enhance model generalization. The VGG19 model is trained end-to-end, with its convolutional layers serving as feature extractors for character recognition. This paper presents a novel approach to Optical Character Recognition (OCR) using the VGG19 convolutional neural network (CNN). OCR is a fundamental technology that converts printed or handwritten text into digital format, facilitating document digitization and information retrieval. The proposed method leverages the hierarchical features learned by VGG19 to accurately extract textual information from images. This study has conducted experiments using publicly available datasets, achieving significant improvements in both training and test accuracy across epochs. Specifically, the proposed model has achieved a training accuracy of 94.34% and a test accuracy of 94.96% after ten epochs of training. Furthermore, we observed a consistent decrease in both training and test loss throughout the training process, indicating effective convergence and refinement of the model parameters. These results demonstrate the efficacy of the VGG19-based OCR model in accurately recognizing characters from diverse input images, highlighting its potential for various real-world applications such as document digitization, augmented reality, and accessibility tools.

A Gaussian Process Upsampling Model for Improvements in Optical Character Recognition

Optimizing the Quality of Fourier Single-Pixel Imaging Via Generative Adversarial Network

Unknown-box Approximation to Improve Optical Character Recognition Performance

Postprocessing Algorithm for the Optical Recognition of Degraded Characters

Improving Text Image Resolution Using a Deep Generative Adversarial Network for Optical Character Recognition

3D Rendering Framework for Data Augmentation in Optical Character Recognition

Point Cloud Upsampling Algorithm: A Systematic Review

A Novel Pipeline for Improving Optical Character Recognition through Post-processing Using Natural Language Processing

UPOCR: Towards Unified Pixel-Level OCR Interface

OCR accuracy improvement on document images through a novel pre-processing approach

General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model

GaussianPU: A Hybrid 2D-3D Upsampling Framework for Enhancing Color Point Clouds via 3D Gaussian Splatting

Survey of Post-OCR Processing Approaches

PixelGaussian: Generalizable 3D Gaussian Reconstruction from Arbitrary Views

Ensemble Model of Attention Mechanism-Based DCGAN and Autoencoder for Noised OCR Classification

A Deep Learning-Based Pre-Trained VGG19 Model for Optical Character Recognition

Advancing Post-OCR Correction: A Comparative Study of Synthetic Data

Iterative Projection Reconstruction for Fast and Efficient Image Upsampling

Efficient, Lexicon-Free OCR using Deep Learning

Improving Offline Gurmukhi Character Recognition: A Comparative Study of Feature Selection Techniques