Geo-Distinctive Visual Element Matching for Location Estimation of Images

Abstract

Abstract-We propose an image representation and matching approach that substantially improves visual-based location estimation for images. The main novelty of the approach, called distinctive visual element matching (DVEM), is its use of representations that are specific to the query image whose location is being predicted. These representations are based on visual element clouds, which robustly capture the connection between the query and visual evidence from candidate locations. We then maximize the influence of visual elements that are geo-distinctive because they do not occur in images taken at many other locations. We carry out experiments and analysis for both geo-constrained and geo-unconstrained location estimation cases using two large-scale, publicly-available datasets: the San Francisco Landmark dataset with 1.06 million street-view images and the MediaEval '15 Placing Task dataset with 5.6 million geotagged images from Flickr. We present examples that illustrate the highly-transparent mechanics of the approach, which are based on common sense observations about the visual patterns in image collections. Our results show that the proposed method delivers a considerable performance improvement compared to the state of the art.

[Retracted] Research on the Application of Multimedia Elements in Visual Communication Art under the Internet Background

“…is publishing method displays multimedia elements such as text, pictures, sound effects, and animations, which are the main elements of information exchange through the terminal screen, giving people convenient access to information and communication [8]. In terms of physical properties, web pages are computer-based information carriers and a collection of thousands of networks worldwide [9]. From its social role, people will naturally regard the network as the fourth media after newspapers, radio, and television.…”

Geotagging Text Content With Language Models and Feature Mining

“…Text-based approaches are classified into two broad categories: geoparsing and Language Model-based (LM). Geotagging approaches based on the visual content of images, such as the ones by Hayes et al [16], [17], Lin et al [35], Weyand et al [54] and Li et al [32], offer another interesting alternative solution to the problem, which is, however, beyond the scope of this paper. Similarly, multimodal approaches that combine both text and visual content to produce location estimates, such as the ones by Crandall et al [9], Kelm et al [20], Trevisiol et al [49] and Cao et al [3] are not further considered here.…”

The Geo-Privacy Bonus of Popular Photo Enhancements

“…Here, we focus on a state-of-the-art GLE system called Distinctive Visual Element Matching approach, which was proposed by Li et al [21]. Given a query image, a geo-location is predicted based on the evidence collected from images that have sufficient visual similarity to the query image and are also geographically close.…”

“…Each cell is considered a class, and is used to train a CNN classifier. Inspired by the result reported in [31], we trained a deep network and compared with the retrieval based approach by Li et al [21].…”

“…Next we describe the GLE systems that were used for the experiments: DVEM, the search approach based on geometric visualfeature matching and a deep convolutional neural network based approach. [21] is a search-based approach to the prediction of the geo-location of an image. This algorithm uses local image features (upright SIFT) for retrieval, and then creates geo-clusters in the list of image search results.…”