Evolutionary Data Purification for Social Media Classification

James, Stuart; Collomosse, John

Abstract

We present a novel algorithm for the semantic labeling of photographs shared via social media. Such imagery is diverse, exhibiting high intra-class variation that demands large training data volumes to learn representative classifiers. Unfortunately image annotation at scale is noisy resulting in errors in the training corpus that confound classifier accuracy. We show how evolutionary algorithms may be applied to select a 'purified' subset of the training corpus to optimize classifier performance. We demonstrate our approach over a variety of image descriptors (including deeply learned features) and support vector machines.

Citation

James, S., & Collomosse, J. (2016). Evolutionary Data Purification for Social Media Classification. In Proceedings of the 2016 23rd International Conference on Pattern Recognition (ICPR). https://doi.org/10.1109/ICPR.2016.7900039

• Conference Name: 2016 23rd International Conference on Pattern Recognition (ICPR)
• Conference Location: Cancun, Mexico
• Start Date: Dec 4, 2016
• End Date: Dec 8, 2016
• Online Publication Date: Apr 24, 2017
• Publication Date: 2016
• Deposit Date: Dec 13, 2023
• Publisher: Institute of Electrical and Electronics Engineers
• Book Title: Proceedings of the 2016 23rd International Conference on Pattern Recognition (ICPR)
• ISBN: 9781509048489
• DOI: https://doi.org/10.1109/ICPR.2016.7900039
• Public URL: https://durham-repository.worktribe.com/output/2024587