A sequential algorithm for training text classifiers
SIGIR '94 Proceedings of the 17th annual international ACM SIGIR conference on Research and development in information retrieval
Content-Based Image Retrieval at the End of the Early Years
IEEE Transactions on Pattern Analysis and Machine Intelligence
Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond
Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond
Object Recognition from Local Scale-Invariant Features
ICCV '99 Proceedings of the International Conference on Computer Vision-Volume 2 - Volume 2
Pattern Classification (2nd Edition)
Pattern Classification (2nd Edition)
Learning Object Categories from Google"s Image Search
ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision - Volume 2
Proceedings of the 13th annual ACM international conference on Multimedia
Probabilistic web image gathering
Proceedings of the 7th ACM SIGMM international workshop on Multimedia information retrieval
To search or to label?: predicting the performance of search-based automatic image classifiers
MIR '06 Proceedings of the 8th ACM international workshop on Multimedia information retrieval
Automatic video annotation by semi-supervised learning with kernel density estimation
MULTIMEDIA '06 Proceedings of the 14th annual ACM international conference on Multimedia
Evaluation of active learning strategies for video indexing
Image Communication
Video diver: generic video indexing with diverse features
Proceedings of the international workshop on Workshop on multimedia information retrieval
Identifying relevant frames in weakly labeled videos for training concept detectors
CIVR '08 Proceedings of the 2008 international conference on Content-based image and video retrieval
A comparison of color features for visual concept classification
CIVR '08 Proceedings of the 2008 international conference on Content-based image and video retrieval
A novel region-based approach to visual concept modeling using web images
MM '08 Proceedings of the 16th ACM international conference on Multimedia
Learning tag relevance by neighbor voting for social image retrieval
MIR '08 Proceedings of the 1st ACM international conference on Multimedia information retrieval
Solving the label resolution problem in supervised video content classification
MIR '08 Proceedings of the 1st ACM international conference on Multimedia information retrieval
Foundations and Trends in Information Retrieval
Can social tagged images aid concept-based video search?
ICME'09 Proceedings of the 2009 IEEE international conference on Multimedia and Expo
Video corpus annotation using active learning
ECIR'08 Proceedings of the IR research, 30th European conference on Advances in information retrieval
Automated image annotation using global features and robust nonparametric density estimation
CIVR'05 Proceedings of the 4th international conference on Image and Video Retrieval
IEEE Transactions on Pattern Analysis and Machine Intelligence
Automatic concept-to-query mapping for web-based concept detector training
MM '11 Proceedings of the 19th ACM international conference on Multimedia
Fusing concept detection and geo context for visual search
Proceedings of the 2nd ACM International Conference on Multimedia Retrieval
Hi-index | 0.00 |
We address the challenge of training visual concept detectors on web video as available from portals such as YouTube. In contrast to high-quality but small manually acquired training sets, this setup permits us to scale up concept detection to very large training sets and concept vocabularies. On the downside, web tags are only weak indicators of concept presence, and web video training data contains lots of non-relevant content. So far, there are two general strategies to overcome this label noise problem, both targeted at discarding non-relevant training content: (1) a manual refinement supported by active learning sample selection, (2) an automatic refinement using relevance filtering. In this paper, we present a highly efficient approach combining these two strategies in an interleaved setup: manually refined samples are directly used to improve relevance filtering, which again provides a good basis for the next active learning sample selection. Our results demonstrate that the proposed combination -- called active relevance filtering -- outperforms both a purely automatic filtering and a manual one based on active learning. For example, by using 50 manual labels per concept, an improvement of 5% over an automatic filtering is achieved, and 6% over active learning. By annotating only 25% of weak positive samples in the training set, a performance comparable to training on ground truth labels is reached.