L1 Graph Based Sparse Model for Label De-noising

Xiaobin Chang, Tao Xiang and Timothy Hospedales

Abstract

The abundant images and user-provided tags available on social media websites provide an intriguing opportunity to scale vision problems beyond the limits imposed by manual dataset collection and annotation. However, exploiting user-tagged data in practice is challenging since it contains many noisy (incorrect and missing) labels. In this work, we propose a novel robust graph-based approach for label de-noising. Specifically, the proposed model is built upon (i) label smoothing via a visual similarity graph in a form of $L_{1}$ graph regulariser, which is more robust against visual outliers than the conventional $L_{2}$ regulariser, and (ii) explicitly modelling the label noise pattern, which helps to further improve de-noising performance. An efficient algorithm is formulated to optimise the proposed model, which contains multiple robust $L_1$ terms in its objective function and is thus non-trivial to optimise. We demonstrate our model's superior de-noising performance across the spectrum of problems from multi-class with label noise to real social media data with more complex multi-label structured label noise patterns.

Session

Recognition and Physics-based vision

Files

Extended Abstract (PDF, 915K)
Paper (PDF, 1M)

DOI

10.5244/C.30.74
https://dx.doi.org/10.5244/C.30.74

Citation

Bibtex

        @inproceedings{BMVC2016_74,
        	title={L1 Graph Based Sparse Model for Label De-noising},
        	author={Xiaobin Chang, Tao Xiang and Timothy Hospedales},
        	year={2016},
        	month={September},
        	pages={74.1-74.12},
        	articleno={74},
        	numpages={12},
        	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
        	publisher={BMVA Press},
        	editor={Richard C. Wilson, Edwin R. Hancock and William A. P. Smith},
        	doi={10.5244/C.30.74},
        	isbn={1-901725-59-6},
        	url={https://dx.doi.org/10.5244/C.30.74}
        }