Thanks to @jovialjoy, I& #39;m reading through this #FairFace paper. There& #39;s a lot to unpack but I just want to point out a couple of underlying fundamental problems with respect to the way these authors think about gender. https://openaccess.thecvf.com/content/WACV2021/papers/Karkkainen_FairFace_Face_Attribute_Dataset_for_Balanced_Race_Gender_and_Age_WACV_2021_paper.pdf">https://openaccess.thecvf.com/content/W...
The authors developed a new face image dataset and trained a model that they claim performs better than previous models on & #39;race, gender, and age classification.& #39; However, even putting aside (for now) ethical problems, they made a number of fundamentally incorrect assumptions:
False assumption 1: there are only two genders.

The authors don& #39;t even provide a caveat, such as & #39;for the purposes of this study we reduced gender to a binary.& #39;

I& #39;m not going to waste time here with this one other than to say, look it up. https://en.wikipedia.org/wiki/Gender ">https://en.wikipedia.org/wiki/Gend...
False assumption 2: humans can correctly classify gender from photos (to provide & #39;ground truth& #39; for model testing). They took a & #39;best 2 out of 3 Turker guesses& #39; approach.
False assumption 3: humans can correctly classify gender from photos ACROSS ALL RACES, ETHNICITIES, AGES, & GENDERS with equal error rates.

This is especially problematic since it undermines the whole point of the paper.
False assumption 4: it& #39;s OK for either humans or machines to classify people& #39;s gender without consent.

It& #39;s not OK to do this. Don& #39;t do it. You need to ask people if you want to know their gender, and get informed consent if you want to use that information to train your model.
False assumption 5: skin = race. It does not. Computer Scientists: Get It Through Your Head. IT DOESN& #39;T. No phenotypical characteristic is a 1-1 match with race because race is not biology. Look it up. https://en.wikipedia.org/wiki/Race_(human_categorization)">https://en.wikipedia.org/wiki/Race...
False assumption 6: "Latino is often treated as an ethnicity, but we consider Latino a race, which can be judged from the facial appearance."

My friends, Latino is not a & #39;race,& #39; ESPECIALLY in this context (computer vision).
Here: https://www.vox.com/2016/8/28/12658908/latino-hispanic-race-ethnicity-explained">https://www.vox.com/2016/8/28...
I have to get back to work now so the thread is done, but there& #39;s a lot more to discuss in this paper. Feel free to do so in replies and I& #39;ll check back in later to RT.
You can follow @schock.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: