Tamil Handwritten Text Recognition using Deep Learning Technique

Text recognition of handwritten Tamil words is a tedious process as the Tamil language contains a most of the characters are similar to each another as well as with complicated structures. A total of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{3 1 3}$</tex> characters are produced by the combination of 156 characters. The majority of the characters have curves, loops, and strokes, which makes the task of recognition difficult. When a character is incorrectly recognized, the word's meaning is modified. A Convolutional Recurrent Neural Network (CRNN) was developed to solve this challenge. This approach efficiently combines convolutional layers for extracting features with recurrent layers for modeling, allowing the system to accurately capture the nuances of Tamil script. The method was designed to capture the long-range dependencies that recognize the character without any errors and acquire a high level of accuracy, precision, recall, and F1 score.