top of page

An Analysis of Energy Requirement for Computer Vision Algorithms

Reference Type: 

Conference Paper

Edelman, Daniel, Siddharth Samsi, Joseph McDonald, Adam Michaleas, and Vijay Gadepally. 2023. “An Analysis of Energy Requirement for Computer Vision Algorithms.” In 2023 IEEE High Performance Extreme Computing Conference (HPEC), 1–7.

The energy requirements of neural network learning are growing at a rapid rate. Increased energy demands have caused a global need to seek ways to improve energy efficiency of neural network learning. This paper aims to establish a baseline on how adjusting basic parameters can affect energy consumption in neural network learning on Computer Vision tasks. In this article, we catalog the effects of various adjustments, from simple batch size adjustments to more complicated hardware settings (e.g., power capping). Based on our characterizations, we have found numerous avenues to adjust computer vision algorithm energy expenditure. For example, switching from a single precision model to mixed precision training can result in energy reductions of nearly 40 %. dditionally, power capping the Graphical Processing Unit (GPU) can reduce energy cost by an additional 10%.

Download Reference:

Search for the Publication In:

Formatted Reference:

bottom of page