Iterative Reranking of Relevant Images (software)

Resource Type	Date
Software	2015-08-20

Description

Authors: Aniol Lidon (UPC), Marc Bolaños (UB), Markus Seidl (STP), Xavier Giró-i-Nieto (UPC), Petia Radeva (UB) and Matthias Zeppelzauer (STP).

This page contains some of the tools used in the UPC-UB-STP submission to the 2015 MediaEval Retrieving Diverse Images Task.

Relevance CNN

Download (217 MB)

A Relevance convolutional neural network (CNN) was created based on HybridNet [1], a CNN trained with objects from the ImageNet dataset [2] and locations from the Places dataset [1]. HybridNet was fine-tuned in two classes: relevant and irrelevant, as labeled by human annotators [3]. Use Caffe to read this model.

External references

[1] B. Zhou, A. Lapedriza, J. Xiao, A. Torralba, and A. Oliva. Learning deep features for scene recognition using places database. In Advances in Neural Information Processing Systems, pages 487–495, 2014. [Project site]

[2] J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. Imagenet: A large-scale hierarchical image database. In Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on, pages 248–255. IEEE, 2009. [Project site]

[3] B. Ionescu, A. L. Gınsca, B. Boteanu, A. Popescu, M. Lupu, and H. Müller. Retrieving diverse social images at mediaeval 2015: Challenge, dataset and evaluation. In MediaEval 2015 Workshop, Wurzen, Germany, 2015. [Project site]

[4] Y. Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Girshick, S. Guadarrama, and T. Darrell. Caffe: Convolutional architecture for fast feature embedding. In Proceedings of the ACM International Conference on Multimedia, pages 675–678. ACM, 2014. [Project site]