Advancements in Communication and Systems
Face Generation using GAN
Authors: Sakshi Singh, Neha Lalit, Ameesha Roy, Akanksha Chauhan, Konika Rani and Neha Gautam
Publishing Date: 20-01-2024
ISBN: 978-81-955020-7-3
Abstract
With applications ranging from entertainment and art to AI-assisted human- computer interaction, the need for realistic face generation has increased dramatically. By employing a multi-modal system that seamlessly blends generative adversarial networks (GANs), natural language processing (NLP), and an attention mechanism, this study offers a novel approach to solving this issue. The FFHQ annotated dataset is used in this work to extract subtle facial traits from user-provided text descriptions using NLP-based feature vectors. An attention technique is used to further improve the images produced by a Conditional GAN, directing the model to concentrate on text-conditioned regions. The method generates diverse, high-quality face images that closely match user-specified criteria by utilizing multi-GAN capabilities. The efficacy of this approach is demonstrated by the experimental results, which are qualitative with user-generated material and quantitative with measures such as Inception Score. This work advances the state of the art in the field of multimodal face generation by providing a promising path for it.
Keywords
Natural language processing (NLP), vector encoding, Generative adversarial networks (GANs), Machine Learning (ML), Multi-GAN, Conditional GAN, Deep Convolutional Generative Adversarial Networks (DCGAN), Attention Mechanism.
Cite as
Sakshi Singh, Neha Lalit, Ameesha Roy, Akanksha Chauhan, Konika Rani and Neha Gautam, "Face Generation using GAN", In: Ashish Kumar Tripathi and Vivek Shrivastava (eds), Advancements in Communication and Systems, SCRS, India, 2024, pp. 115-120. https://doi.org/10.56155/978-81-955020-7-3-10