Image Generation using Variational Autoencoders
Purnima Sai Koumudi Panguluri1, Kishore Kumar Kamarajugadda2
1Purnima Sai Koumudi Panguluri, Department of CSE, Faculty of Science and Technology, ICFAI Foundation for Higher Education, Hyderabad, India.
2Kishore Kumar Kamarajugadda, Department of ECE, Faculty of Science and Technology, ICFAI Foundation for Higher Education, Hyderabad, India.
Manuscript received on February 10, 2020. | Revised Manuscript received on February 23, 2020. | Manuscript published on March 10, 2020. | PP: 517-520 | Volume-9 Issue-5, March 2020. | Retrieval Number: E2480039520/2020©BEIESP | DOI: 10.35940/ijitee.E2480.039520
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The proposed system generates new images from the existing images using variational autoencoders. The autoencoder aims to map the input image to a multivariate normal distribution in the latent space. Variational autoencoder transforms input image into a remarkable output by reducing the reconstruction and KL divergence losses. The primary advantage of implementing variational autoencoder over the other autoencoders is that it follows a specific probability distribution called Gaussian distribution and results in generating high quality images.
Keywords: Variational Autoencoders, Principal Component Analysis, Orthogonal Transformation, KL Divergence, Encoder and Decoder.
Scope of the Article: Image Analysis and Processing