English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Color vision is essential for human visual perception, but its impact onmachine perception is still underexplored. There has been an intensified demandfor understanding its role in machine perception for safety-critical tasks suchas assistive driving and surgery but lacking suitable datasets. To fill thisgap, we curate multipurpose datasets ColorSense, by collecting 110,000non-trivial human annotations of foreground and background color labels frompopular visual recognition benchmarks. To investigate the impact of colorvision on machine perception, we assign each image a color discrimination levelbased on its dominant foreground and background colors and use it to study theimpact of color vision on machine perception. We validate the use of ourdatasets by demonstrating that the level of color discrimination has adominating effect on the performance of mainstream machine perception models.Specifically, we examine the perception ability of machine vision byconsidering key factors such as model architecture, training objective, modelsize, training data, and task complexity. Furthermore, to investigate how colorand environmental factors affect the robustness of visual recognition inmachine perception, we integrate our ColorSense datasets with image corruptionsand perform a more comprehensive visual perception evaluation. Our findingssuggest that object recognition tasks such as classification and localizationare susceptible to color vision bias, especially for high-stakes cases such asvehicle classes, and advanced mitigation techniques such as data augmentationand so on only give marginal improvement. Our analyses highlight the need fornew approaches toward the performance evaluation of machine perception modelsin real-world applications. Lastly, we present various potential applicationsof ColorSense such as studying spurious correlations.

ColorSense: A Study on Color Vision in Machine Visual Recognition