Consistent Display of Clemson Brand Colors Using Artificial Intelligence

Details

Document ID: 
200074
Author(s): 
Erica Walker, Hudson Smith, Emma Mayes, John Paul Lineberger, Michelle Mayer, and Andrew Sanborn
Year: 
2020
Pages: 
10

Pricing

Digital, Non-Member: 
$20.00
Photo, Member: 
$15.00
Photo, Non-Member: 
$30.00

Abstract

When watching a broadcast of an athletic event on a digital display, fans notice when there are inconsistencies in brand colors. Brand colors are a valuable asset built through consistency and repetition. The accurate display of color is impacted by factors including the capture source(s), display screen technology, and ambient lighting which can change drastically during a live event due to changes in time of day and weather. Environmental shifts require adjustments to the video feed to maintain visual consistency. In the current workflow, this is done manually by a technician who color-corrects up to two dozen incoming camera feeds in realtime. The primary question of this research is: Can ColorNet, a neural networkbased algorithm, automatically color correct for accurate and consistent display of brand colors in real-time video without impacting non-targeted colors? ColorNet is a patent-pending artificial intelligence (AI) technology that applies a machine learning model to adjust each video frame pixel-by-pixel to produce a colorcorrected video output. For this study, ColorNet is demonstrated using Clemson University's brand color, Pantone 165 (orange). The model was trained using a collection of corresponding original and color-corrected frames from Clemson athletic events. Manual color correction was completed using Adobe Premiere Pro to produce this dataset. The current model is able to adjust only the targeted brand colors without shifting surrounding colors in the frame, generating localized corrections while adjusting automatically to changes in lighting; this is validated through analysis of the impact of ColorNet adjustments on the full color spectrum. The production viability of ColorNet was demonstrated through alpha and beta tests, and a Delta E (dE) color distance analysis was used to determine if the djustment for the targeted colors was within a reasonable tolerance of the brand specification. Further progress is being made to expand the model to additional brand colors and explore applications beyond sports broadcasting.

Return To Search Results

Search Again

TAGA Papers Order Form