9th International Conference on ICT for Sustainable Development

Authors - Punam Dawgotra
Abstract - This systematic study presents a comprehensive analysis of datasets utilized in deep learning-based face generation research. Datasets in face generation using deep learning serve as the foundation upon which models learn to generate realistic human faces. This systematic survey explores various datasets comprising of human face images collected from various sources like Internet, social media platforms or photo gallery portals like Flickr and provides insights into their characteristics, such as size, resolution, annotations, and applications. By examining the evolution of datasets over time, from early benchmarks to recent high-resolution collections, this survey highlights the importance of dataset quality and diversity in implementing deep learning models in face generation techniques. It discusses various widely popular deep learning algorithms like GAN based DCGAN, CycleGAN, CGAN, StyleGAN, PGGAN, StyleGAN2, AttentionGAN and Autoencoder based VAE Model which are currently being utilized in human face generation. Through this survey, researchers and developers gain valuable insights into the landscape of face generation datasets, facilitating informed decision-making and advancements in the field of face generation applications.