Complex Event Detection using Semantic Saliency and Nearly-Isotonic SVM

Xiaojun Chang, Yi Yang, Eric Xing, Yaoliang Yu
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:1348-1357, 2015.

Abstract

We aim to detect complex events in long Internet videos that may last for hours. A major challenge in this setting is that only a few shots in a long video are relevant to the event of interest while others are irrelevant or even misleading. Instead of indifferently pooling the shots, we first define a novel notion of semantic saliency that assesses the relevance of each shot with the event of interest. We then prioritize the shots according to their saliency scores since shots that are semantically more salient are expected to contribute more to the final event detector. Next, we propose a new isotonic regularizer that is able to exploit the semantic ordering information. The resulting nearly-isotonic SVM classifier exhibits higher discriminative power. Computationally, we develop an efficient implementation using the proximal gradient algorithm, and we prove new, closed-form proximal steps. We conduct extensive experiments on three real-world video datasets and confirm the effectiveness of the proposed approach.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-changa15, title = {Complex Event Detection using Semantic Saliency and Nearly-Isotonic SVM}, author = {Chang, Xiaojun and Yang, Yi and Xing, Eric and Yu, Yaoliang}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {1348--1357}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/changa15.pdf}, url = { http://proceedings.mlr.press/v37/changa15.html }, abstract = {We aim to detect complex events in long Internet videos that may last for hours. A major challenge in this setting is that only a few shots in a long video are relevant to the event of interest while others are irrelevant or even misleading. Instead of indifferently pooling the shots, we first define a novel notion of semantic saliency that assesses the relevance of each shot with the event of interest. We then prioritize the shots according to their saliency scores since shots that are semantically more salient are expected to contribute more to the final event detector. Next, we propose a new isotonic regularizer that is able to exploit the semantic ordering information. The resulting nearly-isotonic SVM classifier exhibits higher discriminative power. Computationally, we develop an efficient implementation using the proximal gradient algorithm, and we prove new, closed-form proximal steps. We conduct extensive experiments on three real-world video datasets and confirm the effectiveness of the proposed approach.} }
Endnote
%0 Conference Paper %T Complex Event Detection using Semantic Saliency and Nearly-Isotonic SVM %A Xiaojun Chang %A Yi Yang %A Eric Xing %A Yaoliang Yu %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-changa15 %I PMLR %P 1348--1357 %U http://proceedings.mlr.press/v37/changa15.html %V 37 %X We aim to detect complex events in long Internet videos that may last for hours. A major challenge in this setting is that only a few shots in a long video are relevant to the event of interest while others are irrelevant or even misleading. Instead of indifferently pooling the shots, we first define a novel notion of semantic saliency that assesses the relevance of each shot with the event of interest. We then prioritize the shots according to their saliency scores since shots that are semantically more salient are expected to contribute more to the final event detector. Next, we propose a new isotonic regularizer that is able to exploit the semantic ordering information. The resulting nearly-isotonic SVM classifier exhibits higher discriminative power. Computationally, we develop an efficient implementation using the proximal gradient algorithm, and we prove new, closed-form proximal steps. We conduct extensive experiments on three real-world video datasets and confirm the effectiveness of the proposed approach.
RIS
TY - CPAPER TI - Complex Event Detection using Semantic Saliency and Nearly-Isotonic SVM AU - Xiaojun Chang AU - Yi Yang AU - Eric Xing AU - Yaoliang Yu BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-changa15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 1348 EP - 1357 L1 - http://proceedings.mlr.press/v37/changa15.pdf UR - http://proceedings.mlr.press/v37/changa15.html AB - We aim to detect complex events in long Internet videos that may last for hours. A major challenge in this setting is that only a few shots in a long video are relevant to the event of interest while others are irrelevant or even misleading. Instead of indifferently pooling the shots, we first define a novel notion of semantic saliency that assesses the relevance of each shot with the event of interest. We then prioritize the shots according to their saliency scores since shots that are semantically more salient are expected to contribute more to the final event detector. Next, we propose a new isotonic regularizer that is able to exploit the semantic ordering information. The resulting nearly-isotonic SVM classifier exhibits higher discriminative power. Computationally, we develop an efficient implementation using the proximal gradient algorithm, and we prove new, closed-form proximal steps. We conduct extensive experiments on three real-world video datasets and confirm the effectiveness of the proposed approach. ER -
APA
Chang, X., Yang, Y., Xing, E. & Yu, Y.. (2015). Complex Event Detection using Semantic Saliency and Nearly-Isotonic SVM. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:1348-1357 Available from http://proceedings.mlr.press/v37/changa15.html .

Related Material