Fast shared boosting: Application to large-scale visual concept detection

This work addresses the problem of large-scale visual concept detection. Visual concepts are usually learned from an annotated image or video database with a machine learning algorithm, posing this problem as a multiclass supervised learning task. Some practical issues appear when the number of concept grows, in particular when one aims at developing applications for real users, restricting the constraints in terms of available memory and computing time (both for learning and testing). To cope with these issues, we propose in this article to use a multiclass boosting with feature sharing algorithm and reduce its computational complexity with a set of efficient improvements. This makes our algorithm able to handle a problem of classification with many classes in a reasonable time. The relevance of our algorithm is evaluated in the context of information retrieval, on the benchmark proposed into the ImageCLEF international evaluation campaign and shows competitive results.