content-based image retrieval

Recommender systems, with the ability to recognize the user and predict his priorities, sift through the information that is likely to be of interest to the user from the large volume of data and save the user's time and energy by recommending them to the user. On the other hand, these systems, with the ability to analyze and store the user's past behaviors, also infer existing services and information that the user has not paid attention to but is probably interested in and provide interesting results to users in the form of recommendations. While in the vast majority of works done in the field of recommender systems, each item is displayed only with numerical or string features, one of the features that has a significant impact on determining the level of desirability of items from the users' point of view is their images. This is more important for products such as clothing, jewelry, etc., which are mostly accepted by customers due to their physical appearance. For this purpose, this paper aims to present a combined method to increase the efficiency of these systems using collaborative filtering techniques and the content of the images of the items. In this paper, first, using image processing techniques, the visual features of the items are extracted, then by selecting the appropriate similarity criterion and using the technical specifications of the items and the individual characteristics of the users, the users and similar items will be segmented. After that, using this information and utilizing the collaborative filtering technique, the best suggestions that are most similar to the users' tastes will be presented. Experimental results show that the proposed method has been able to demonstrate appropriate efficiency.

Content based image retrieval (CBIR) applies machine vision techniques to extract similar images for a given query image. The main challenge of CBIR is the semantic gap between low level pixel and segment based features and high-level concepts in the image. An approach towards reducing this gaps is to use high level region and object based features. However, the low-level features describe image details and enforce between image discriminations. Accordingly, it is expected that the use of both feature types will lead to better results. This paper tries to reduce the mentioned gap by combining decision results at four granularities, namely pixel, region, object, and concept levels. Pixel level retrieval adopts SIFT features and local binary patterns. Region level subsystem partitions the image into a set of segments and extracts their color and texture features using hue descriptor and Gabor filters for subsequent processing. AlexNet convolutional neural network is employed for object based retrieval. Word2vec embedding is used for concept level retrieval that exploits conceptual relations between objects to enhance the retrieval results. Experiments over Wang and GHIM datasets confirm the feasibility of the proposed combination and conclude that it improves overall performance of the retrieval system.