Author: Azizian, Bardia
In this thesis, we study different approaches to visual data coding for machines and develop new methods to address some issues in this area. We mainly focus on the Image and Video signals processed by a Deep Neural Network (DNN)-based computer vision model. Our proposed methods are designed to be utilized in DNN-based machines deployed collaboratively on the edge and cloud. This framework is called Collaborative Intelligence (CI), in which a DNN model is split into two parts such that the edge device runs the first few layers, and the remaining layers are executed on the cloud. To that end, the intermediate feature tensors need to be coded and transferred to the cloud. This research explicitly attempts to provide solutions for efficient coding of these tensors, considering challenges such as motion estimation and compensation for videos in the latent space, and privacy for images.
Copyright is held by the author(s).
This thesis may be printed or downloaded for non-commercial research and scholarly purposes.
Supervisor or Senior Supervisor
Thesis advisor: bajic, ivan
Member of collection