Recommending Fashion Pattern Clothes Based on Collaborative Filtering and Recurrent Neural Network - A Maximum Likelihood Classification Approach

Fashion accoutrements recommendation system is often problematic area in which pattern clothes configuration with images is challenging and it needs creativity as well. Combining items like jewelry, bag, pants, cap, dress, shoes with appearance. In fashion websites, popular or high-quality fashion clothes in India are designed by fashion experts and followed by large spectators. Recurrent neural network has been proposed in this paper. The center of the proposed motorized configuration system is to score fashion pattern clothes candidates based on the appearances and metadata. We propose to influence pattern clothes popularity on fashion-oriented websites to supervise the fashion acceptance score component. The fashion acceptance score part is a avant-garded multiple modal multiple occurrence deep learning system that evaluates instance artistic and set compatibility simultaneously. In order to coach and assess the proposed configuration system, we have collected a large-scale fashion pattern clothes dataset with 208K clothes and 668K fashion items from Fancy data. Although the fashion pattern clothes fashion acceptance score and configuration is rather challenging, we have achieved an AUC of 90% for the fashion acceptance score component, and an accuracy of 83% for a constrained configuration task.