Mining weakly labeled web facial images for search-based face annotation

Abstract

In this paper, we investigate a search-based face annotation framework by mining weakly labeled facial images that are freely available on the internet. A key component of such a search-based annotation paradigm is to build a database of facial images with accurate labels. This is however challenging since facial images on the WWW are often noisy and incomplete. To improve the label quality of raw web facial images, we propose an effective Unsupervised Label Refinement (ULR) approach for refining the labels of web facial images by exploring machine learning techniques. We develop effective optimization algorithms to solve the large-scale learning tasks efficiently, and conduct an extensive empirical study on a web facial image database with 400 persons and 40,000 web facial images. Encouraging results showed that the proposed ULR technique can significantly boost the performance of the promising search-based face annotation scheme.

An overview of content-based image retrieval techniques (CBIR)

“…Website owners made a name consisting of 2.45 million distinct image that has 421 436 names and has order of large magnitude scales over previous data sets. Collecting and labeling of such a large dataset, is a big challenge in the current multimedia search methods as well [24].…”

Image mining framework and techniques: a review

“…This issue demanded a simpler approach for image mining (Jain et al, 2013). Thus, the technique of image mining (Banda et al, 2014;Chen and Mei, 2014) was divided into four important phases: image pre-processing, feature extraction, conversion of image database to transaction database, and applying association rule mining (Wang et al, 2014;Herold et al, 2011;Khodaskar and Ladhake, 2014) to this transaction database. The proposed new association rule algorithm (Deshpande, 2010) reduced the number of scans for a priori algorithm.…”

Protein Function Prediction with Incomplete Annotations

“…In this scenario, given an annotated instance, it is unknown whether the annotations are complete or partial. Some approaches developed to replenish the missing (or noisy) labels in the single label case [38], and few methods are developed for multi-label learning scenarios.…”