Random Patterns Clothing Image Retrieval using Convolutional Neural Network

In this paper we present supervised semantic preserving deep hashing (SSDH) based random patterns clothing image retrieval framework. The dense existence of random patterns in Pakistani women's clothing (PWC) images makes them different and challenging as compared to the western women's clothing images. Owing to the diversity of PWC, we propose an efficient representation of their random patterns in terms of the basic patterns found in western women's clothing. We present a retrieval framework for PWC consisting of two stages which are attribute selection and binary coding of features. Regarding the attribute selection, we use the texture that better describes the dense patterns of PW C, and the binary coding of features is performed by using SSD H technique. In order to evaluate the performance of the proposed PW C retrieval framework we present a dataset of PWC images. Experimental evaluation shows that the proposed representation is capable of dealing with the high variations of PWC random patterns effectively.