Adaptive Convolutional Neural Network Based Handwritten Tamil Character Recognition System

Tamil Handwritten Character Recognition (THCR) is a pivotal task in preserving and digitizing the cultural and linguistic heritage embedded in the Tamil script. This paper introduces an innovative approach to THCR using an Adaptive Convolutional Neural Network (ACNN). The proposed ACNN dynamically adjusts its convolutional filters during the training process, allowing it to adapt to the intricate variations and complexities present in handwritten Tamil characters. Leveraging a diverse dataset encompassing various writing styles and regional influences, the ACNN demonstrates superior adaptability and recognition accuracy compared to traditional convolutional neural networks. The adaptive nature of the model makes it well-suited for handling the inherent variability in handwritten Tamil script, offering a robust solution for the efficient transcription of diverse and intricate characters into digital format. The experimental results showcase the efficacy of the ACNN in advancing the state-of-the-art in THCR, providing a promising avenue for the preservation and digitization of Tamil cultural and linguistic heritage.