Generation-Of-Nepali-Hand-written-letter-using-Generative-Adverserial-Network
The Generative modeling became recently one of the most important field in deep learning, and the Generative Adversarial Networks is the new research line in this field, the GANs have proven their ability to generate a high resolution of images and achieved remarkable success for computer vision in general, by the same way, we assume that this tool will be able to generate accurate Nepali letters. Our model uses Deep Convolutional Generative Adversarial Network architecture to generate the Nepali Handwritten Images after training on real-life datasets. It uses padding and strides to get better and good result than normal Generative network. Real- life datasets contains 36 different classes of Handwritten Images of size 64x64 which are feed to our model. We defined the Convolutional Neural Network Architecture for Generator and Discriminator with neccessary hyperparameters and then trained the model to generate output. The expected benefit of this work explores the potential to serve the Nepali calligraphy. As an impact for further studies in the future, this work may provide an insight into the possibility of generating new kinds of Nepali calligraphy.
