Dress me up!: content-based clothing image retrieval

Abstract

It is sometimes difficult for a person to find a matching outfit in a clothing store. It would be beneficial if there is a mechanism which allows a person to provide an example of clothes that they are looking for in the form of an image instead of text description and they will be provided with similar items that the store have. In this paper, we propose a framework for content-based clothing image retrieval that allows searching and retrieving of similar clothes based on colour and shape features. We first convert the RGB image to HSV colour space. Colour histogram is then computed for colour representation while the shape representation is obtained from low and high thresholding computation. The similarity between input image and image collection is computed based on Manhattan distance. The proposed framework can aid users in finding the desired and similar clothing images from catalogues effectively without the needs of text-based input. Users can also search for similar clothing based on specific features like colour or shape. For each clothes, the item s descriptions like brand, size, and price are also being displayed. We have conducted retrieval effectiveness experiment and questionnaire-based survey to measure the effectiveness of our framework and users acceptance. We have obtained good results for both evaluation approaches.