Improving accuracy in Text Extraction From images using Region-Based Convolutional Neural Networks algorithm compared to Convolutional Neural Network algorithm

Text extraction is a critical task in data processing and analysis, requiring high accuracy for effective applications. This research employed a publicly accessible Kaggle dataset to assess the performance of two neural network architectures that are Region-Based Convolutional Neural Networks (R-CNN) with that of standard Convolutional Neural Networks (CNN). A total sample of 40 test images were taken and segregated into groups, with 20 samples allocated to each method. Implementation was conducted using Python, and statistical analysis was performed using Clincalc.com. The study set a statistical power of 85%, with alpha level of 0.05 and 0.2 beta level set to ensure a thorough analysis. Performance was assessed based on accuracy, with R-CNN achieving 96.461%, significantly outperforming CNN, which recorded an accuracy of 83.583%. The statistical comparison yielded a two-tailed significance value of p < 0.001, highlighting the superior efficiency of R-CNN in text extraction tasks. These findings underscore the potential of R-CNN as a preferred approach for achieving higher accuracy in text extraction applications.