Enhancing Accuracy in Handwritten Text Recognition with Convolutional Recurrent Neural Network and Data Augmentation Techniques

Handwriting recognition is a challenging research area in pattern recognition and image processing that has various applications, such as reading assistance for the blind and digitizing handwritten documents. This study proposes using a convolutional recurrent neural network (CRNN) architecture for handwritten text recognition without the need for segmentation, allowing for the image to be processed in one step. To improve the quality of results and the model, the study employs the CTC decoder to handle the unpredictability that arises when processing multiple alphabets and diverse handwriting. Data augmentation techniques are also utilized to increase the efficiency and accuracy of the model. The study provides detailed information on how to overcome these challenges and improve the accuracy of the proposed model.