My answer to "How advanced is current Computer Vision? Can it tell difference between two vaguely different hairstyle of people?"

The original answer can be found here.

If humans can, machines can as well. From what we have seen (we have already done some work on this at ParallelDots ), clearcut cases can be detected well, but the borderline cases aren’t really clear (the reason is that humans cannot tag it well for AI to learn accurately). People tend to mix and match and since there are no definition for hairstyles, collecting a super accurate dataset is not possible.

Let’s say for beard classification :

For example, the following Goatee with mustache style beard is clear and should be easy to classify using an algorithm:

Similarly will be the classification of the Van Dyke style beard here, should be easy to classify using an algorithm:

But, then the cases in the middle are pretty hard to get objectively annotated by humans (due to no fixed definitions) , and hence an algorithm will find them hard to detect:

Can you classify them into one of the above? The second one has sideburns with a combination of goatee/ Van Dyke. What are the exact parameters of a goatee anyway ? People do not stick to rules while keeping hair/beard styles so it cant be exactly analyzed using Computer Vision. An approximate automated analysis is very much possible.