Cursive Handwriting Recognition Using CNN with VGG-16
Anushka Anil Rangari, Swagat Das, D Rajeswari · 2023
Cursive handwritten text recognition is a process of identifying handwritten text from images. The recognition process is difficult due to person's unique style of writing. Pattern recognition is used to classify the data into various categories. This study used CNN with VGG16 model to identify cursive English alphabets and words in a given scanned text document. The first phase is image acquisition, which involves acquiring the scanned picture, normalizing the image, extracting the features from the image, and applying segmentation. Three different pre-processing techniques like data augmentation, image segmentation and image data generator are implemented in this work. CNN with VGG 16 model is employed for recognition. Three kinds of experiments were done with various combinations of pre-processing techniques combined with CNN model. The experimental results indicates that data augmentation pre-processing technique with CNN produced 98.36% training accuracy and 95.1% testing accuracy which is best than other combinations.