Please use this identifier to cite or link to this item:
https://hdl.handle.net/10356/101852
Title: | Learning sparse tag patterns for social image classification | Authors: | Lin, Jie Duan, Ling-Yu Yuan, Junsong Li, Qingyong Luo, Siwei |
Issue Date: | 2012 | Conference: | IEEE International Conference on Image Processing (19th : 2012 : Orlando, Florida, US) | Abstract: | User-generated tags associated with images from social media (e.g., Flickr) provide valuable textual resources for image classification. However, the noisy and huge tag vocabulary heavily degrades the effectiveness and efficiency of state-of-the-art image classification methods that exploited auxiliary web data. To alleviate the problem, we introduce a Sparse Tag Patterns (STP) model to discover sparsity constrained co-occurrence tag patterns from large scale user contributed tags among social data. To fulfill the compactness and discriminability, we formulate STP as a problem of minimizing a quadratic loss function regularized by the bi-layer l1 norm. We treat the learned STP as alternative intermediate semantic image feature and verify its superiority within a search-based image classification framework. Experiments on 240K social images associated with millions of tags have demonstrated encouraging performance of the proposed method compared to the state-of-the-art. | URI: | https://hdl.handle.net/10356/101852 http://hdl.handle.net/10220/12960 |
DOI: | 10.1109/ICIP.2012.6467501 | Schools: | School of Electrical and Electronic Engineering | Fulltext Permission: | open | Fulltext Availability: | With Fulltext |
Appears in Collections: | SCSE Conference Papers |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
Learning Sparse Tag Patterns for Social Image Classification.pdf | 635.3 kB | Adobe PDF | View/Open |
SCOPUSTM
Citations
50
2
Updated on Mar 14, 2024
Page view(s) 10
900
Updated on Mar 18, 2024
Download(s) 20
264
Updated on Mar 18, 2024
Google ScholarTM
Check
Altmetric
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.