An Approach for Tamil Handwritten Recognition using ABC enabled GAN

Tamil is one of the world's oldest living languages. Recognizing Tamil handwritten character is complex because of insufficient images in datasets, different handwriting styles and sizes. Although many explorations have been done in different languages on character recognition, minimal work has been done so far for Tamil handwritten character recognition. Deep learning has recently been of great interest in computer science as well as in recognizing handwritten characters. DNN-based model has been made use of to recognize the handwritten Tamil characters, where CNNs have played a key role in the area of handwritten character recognition. The proposed approach consists of three stages: pre-processing to remove noise from the images; character augmentation for generating images; and classification for recognizing characters. Since there is a scarcity of Tamil handwritten datasets, we have applied GANs for data augmentation. In order to improve the performance of the GAN, Artificial bee colony Optimization algorithm is incorporated with GAN. This algorithm begins with generating a population of solutions that are then iteratively improved to enhance the efficiency of the GAN. Then, optimized weights are applied to the GAN, and then the model's performance is tested using a specific test dataset to check the efficiency.