Dimension Reduction of Deep Network Models for Mobile Computational Plenoptics

NAISS 2023/22-280


NAISS Small Compute

Principal Investigator:

Ali Hassan



Start Date:


End Date:


Primary Classification:

10206: Computer Engineering



Recent advancement in image acquisition devices has enabled users to capture the spatial and angular information of the scene, known as Light Field (LF). With the rise of Deep Learning (DL), many learning-based algorithms are proposed and have achieved significant improvement in the various field of Light Field applications, including disparity estimation, view synthesis and super-resolution. The majority of recent successful DL architectures require millions of parameters. This makes them energy, computation, and memory intensive. As a result, such architectures require long inference time and power-consuming computational resources (e.g., GPU). For these reasons, compression of DL architecture is essential for the training and deployment of such models in practice. In my research, i will be compressing deep learning architecture specially designed for Light Field application. I have to train and evaluate much state-of-the-art deep learning architecture, as well as new proposed architecture. I would like to use HPC2N resources to perform various time-consuming simulations of my research.