Cornell University's Lee Humphreys, Chelsea Butkowski, and Utkarsh Mall uncovered discrepancies in the skin tones of models depicted in online retail images.
The researchers developed a method that quantitatively measured skin lightness and darkness in still photos and videos, to quantify inconsistencies in different depictions of the same model across platforms.
They analyzed 30 images, focusing on the chin and all visible skin, and generated grayscale histograms to visualize pixel intensity and clustering across the range of possible tones.
Although Humphreys found the images encouraging for their ethnic ambiguity, the skin tones were still relatively light.
"What becomes really interesting is that the discrepancy itself becomes problematic, and a potential indicator of photo manipulation or technological bias," Humphreys said.
From Cornell Chronicle
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA