FAQ
What is the goal of this classification task?
Our goal is to determine the underlying racial and ethnic phenotypic classifications of human images in the largest computer vision training databases.
Why should I do this task?
- It's easy.
- It's kinda fun.
- Most importantly, this work may eventually improve the algorithms that underlie many software platforms, and make the world a fairer place for all human beings, regardless of their race, ethnicity, or skin color.
Who are you?
We are a research team that is dedicated to tackling race and prejudice in consumer technology and marketing. This particular project is under the guidance of Broderick Turner (Ph.D., Marketing; Northwestern University), an assistant professor of marketing at Virginia Tech.
What will you do with the data?
Our goal is to make the data public so that any platform or software that is using this training data can update its software to make it fairer. We will also use this data to determine how salience-based image cropping algorithms may be biased.
How long does it take to do a classification task?
Depending on the image, it may take between 3 seconds and 1 minute to do classification.
How many different images are there in this Zooniverse project?
To date, there are 2300 images that are to be classified. The first round includes all the images from the MIT300 saliency benchmark set.
Can I do more than one classification?
Sure. Do one. Do them all. This is how I (Broderick Turner) relax after a long day.
What do you mean by White, Black, East Asian, South Asian or Other Ethnicity?
First, we want to be clear, there is no biological racial classification. However, most people carry a relatively consistent, and cogent mental model for the phenotypic markers (hair texture, eye shape, skin color) for different racial classifications. So, for each race or ethnicity we ask you to identify by the following with examples from the face research lab, London set:
White: racialized classification of people and a skin color specifier, generally used for people of European origin; although the definition can vary depending on context, nationality, and point of view
Black: racialized classification of people and a skin color specifier, generally used for people of African origin; although the definition can vary depending on context, nationality, and point of view
East Asian: racialized and ethnic classification of people and a skin color specifier, generally used for people with origins from China, Japan, Mongolia, North Korea, South Korea, and Taiwan.
South Asian: racialized and ethnic classification of people and a skin color specifier, generally used for people with origins from Afghanistan, Pakistan, India, Nepal, Bhutan, Bangladesh, Maldives, and Sri Lanka.
Other/ Ethnicity Not Listed: These are people who do not clearly fit any of the listed racial or ethnic categories (White, Black, East Asian, or South Asian).
I want to know more about this work. How can I reach you?
You can email Broderick at brodericklt@vt.edu. He is slow to respond, but he will respond.