Comparative Study of Deep Learning Models for Devanagari OCR
Shilpa Kaur Manocha, Piyush Tewari · 2021 International Conference on Smart Generation Computing, Communication and Networking (SMART GENCON) · 2021
Emergence of Deep Learning Approaches have led to major breakthroughs and improvement in image classification tasks. Handwritten character recognition is one such area where the deep learning methods are employed for enhanced performance. This paper reviews the state-of-the-art deep learning methods viz. CNN, ANN, Capsule Network, Residual Network, and Transfer Learning Networks that are pre-trained on the ImageNet dataset— AlexNet, DenseNet, VGG, Inception used for recognition of the handwritten characters of Devanagari script. The paper presents an exhaustive comparison of all mentioned techniques both in terms of qualitative and quantitative analysis. The paper also highlights the challenges, and future directions for Devanagari characters recognition.