Fast Image Tagging

Minmin Chen, Alice Zheng, Kilian Weinberger
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):1274-1282, 2013.

Abstract

Automatic image annotation is a difficult and highly relevant machine learning task. Recent advances have significantly improved the state-of-the-art in retrieval accuracy with algorithms based on nearest neighbor classification in carefully learned metric spaces. But this comes at a price of increased computational complexity during training and testing. We propose FastTag, a novel algorithm that achieves comparable results with two simple linear mappings that are co-regularized in a joint convex loss function. The loss function can be efficiently optimized in closed form updates, which allows us to incorporate a large number of image descriptors cheaply. On several standard real-world benchmark data sets, we demonstrate that FastTag matches the current state-of-the-art in tagging quality, yet reduces the training and testing times by several orders of magnitude and has lower asymptotic complexity.

Cite this Paper

BibTeX
@InProceedings{pmlr-v28-chen13j, title = {Fast Image Tagging}, author = {Chen, Minmin and Zheng, Alice and Weinberger, Kilian}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {1274--1282}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/chen13j.pdf}, url = {https://proceedings.mlr.press/v28/chen13j.html}, abstract = {Automatic image annotation is a difficult and highly relevant machine learning task. Recent advances have significantly improved the state-of-the-art in retrieval accuracy with algorithms based on nearest neighbor classification in carefully learned metric spaces. But this comes at a price of increased computational complexity during training and testing. We propose FastTag, a novel algorithm that achieves comparable results with two simple linear mappings that are co-regularized in a joint convex loss function. The loss function can be efficiently optimized in closed form updates, which allows us to incorporate a large number of image descriptors cheaply. On several standard real-world benchmark data sets, we demonstrate that FastTag matches the current state-of-the-art in tagging quality, yet reduces the training and testing times by several orders of magnitude and has lower asymptotic complexity.} }
Endnote
%0 Conference Paper %T Fast Image Tagging %A Minmin Chen %A Alice Zheng %A Kilian Weinberger %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-chen13j %I PMLR %P 1274--1282 %U https://proceedings.mlr.press/v28/chen13j.html %V 28 %N 3 %X Automatic image annotation is a difficult and highly relevant machine learning task. Recent advances have significantly improved the state-of-the-art in retrieval accuracy with algorithms based on nearest neighbor classification in carefully learned metric spaces. But this comes at a price of increased computational complexity during training and testing. We propose FastTag, a novel algorithm that achieves comparable results with two simple linear mappings that are co-regularized in a joint convex loss function. The loss function can be efficiently optimized in closed form updates, which allows us to incorporate a large number of image descriptors cheaply. On several standard real-world benchmark data sets, we demonstrate that FastTag matches the current state-of-the-art in tagging quality, yet reduces the training and testing times by several orders of magnitude and has lower asymptotic complexity.
RIS
TY - CPAPER TI - Fast Image Tagging AU - Minmin Chen AU - Alice Zheng AU - Kilian Weinberger BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-chen13j PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 1274 EP - 1282 L1 - http://proceedings.mlr.press/v28/chen13j.pdf UR - https://proceedings.mlr.press/v28/chen13j.html AB - Automatic image annotation is a difficult and highly relevant machine learning task. Recent advances have significantly improved the state-of-the-art in retrieval accuracy with algorithms based on nearest neighbor classification in carefully learned metric spaces. But this comes at a price of increased computational complexity during training and testing. We propose FastTag, a novel algorithm that achieves comparable results with two simple linear mappings that are co-regularized in a joint convex loss function. The loss function can be efficiently optimized in closed form updates, which allows us to incorporate a large number of image descriptors cheaply. On several standard real-world benchmark data sets, we demonstrate that FastTag matches the current state-of-the-art in tagging quality, yet reduces the training and testing times by several orders of magnitude and has lower asymptotic complexity. ER -
APA
Chen, M., Zheng, A. & Weinberger, K.. (2013). Fast Image Tagging. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):1274-1282 Available from https://proceedings.mlr.press/v28/chen13j.html.

Related Material