Outfit Preference Prediction

Teaching AI to predict outfit preference is a high-level subjective problem. There are a lot of ways of combating the problem of computer-learned subjective preferences. One of the most effective ways is to reduce it to a lower level problem. In research, we have gone in multiple directions, such as reducing the problem to problems of expert knowledge, image similarity, visual regularized collaborative filtering, image classification and image regression. Finally, we've gotten the best results by reducing the problem to one of image classification. Classes, in this case, were the scores of the outfit preference. For classification learning, we have used the Inception ResNet v2 implementation of the convolutional neural network. The final solution was presented as a two-tier classification with convolutional neural networks.