The visual identity of a person is an important part of human non-verbal communication. The impression that an individual leaves on others is often based on their fashion style. That is precisely the reason that makes it completely valid for people to be concerned about what they wear in order to leave a good impression on others. The question arises - can a computer help with that?
The first step in trying to answer that question is to make the computer able to automatically detect clothing in images. That was the main goal of this project.
In order to solve that particular problem we used the computer vision methods for semantic segmentation, backed by deep machine learning models. In particular, the model that showed the best results was the so-called fully convolutional neural network.
After detecting the clothes, the next crucial step in helping people dress better was to predict how others like their outfit. That is where outfit preference prediction project comes into play.