Sorry, you need to enable JavaScript to visit this website.

Edge-Cloud Collaborative Multimedia Analysis

Citation Author(s):
Ivan V. Bajic
Submitted by:
Ivan Bajic
Last updated:
30 July 2023 - 4:37pm
Document Type:
Document Year:
Ivan V. Bajic

Our world is at the beginning of the technological revolution that promises to transform the way we work, travel, learn, and live, through Artificial Intelligence (AI). While AI models have been making tremendous progress in research labs and overtaking scientific literature in many fields, efforts are now being made to take these models out of the lab and create products around them, which could compete with established technologies in terms of cost, reliability, and user trust, as well as enable new, previously unimagined applications. Foremost among these efforts involves bringing AI “to the edge” by pairing it with the multitude of sensors that is about to cover our world as part of the Internet of Things (IoT) and 5th generation (5G) communication network initiatives.

This tutorial is about edge-cloud collaborative analysis of multimedia signals, which we shall refer to as Collaborative Intelligence (CI). This is a framework in which AI models, developed for multimedia signal analysis, are distributed between the edge devices and the cloud. In CI, typically, the front-end of an AI model is deployed on an edge device, where it performs initial processing and feature computation. These intermediate features are then sent to the cloud, where the back-end of the AI model completes the inference. CI has been shown to have the potential for energy and latency savings compared to the more typical cloud-based or fully edge-based AI model deployment, but it also introduces new challenges, which require new science and engineering principles to be developed in order to achieve optimal designs. In CI, a capacity-limited channel is inserted in the information pathway of an AI model. This necessitates compression of features computed at the edge sub-model, which in turn requires a solid understanding of the structure of the model’s latent space. Errors introduced into features due to channel imperfections would need to be handled at the cloud side in order to perform successful inference. Moreover, issues related to the privacy of transmitted data need to be addressed.

0 users have voted: