Dynamically Visual Disambiguation of Keyword-based Image Search

TL;DR

This paper introduces an adaptive multi-model framework that dynamically disambiguates keywords in image search results by employing saliency-guided deep learning, improving the accuracy of web-based image retrieval.

Contribution

It proposes a novel adaptive framework combining dynamic query selection and saliency-guided deep learning for visual disambiguation in keyword-based image search.

Findings

Outperforms existing methods in visual disambiguation tasks

Effectively adapts to dynamic changes in search results

Demonstrates superior accuracy in extensive experiments

Abstract

Due to the high cost of manual annotation, learning directly from the web has attracted broad attention. One issue that limits their performance is the problem of visual polysemy. To address this issue, we present an adaptive multi-model framework that resolves polysemy by visual disambiguation. Compared to existing methods, the primary advantage of our approach lies in that our approach can adapt to the dynamic changes in the search results. Our proposed framework consists of two major steps: we first discover and dynamically select the text queries according to the image search results, then we employ the proposed saliency-guided deep multi-instance learning network to remove outliers and learn classification models for visual disambiguation. Extensive experiments demonstrate the superiority of our proposed approach.