好文作者面授招–20150805

【15-24期VALSE Webinar活动】

报告嘉宾1:徐仲文(悉尼科技大学)
主持人:张利军(南京大学)
报告题目:A Discriminative CNN Video Representation for Event Detection [Slides]
报告时间:2015年8月5日晚20:00(北京时间)
文章信息:
[1] Z. Xu, Y. Yang and A. G. Hauptmann, A Discriminative CNN Video Representation for Event Detection, In CVPR, 2015
报告摘要:In this paper, we propose a discriminative video representation for event detection over a large scale video dataset when only limited hardware resources are available. The focus of this paper is to effectively leverage deep Convolutional Neural Networks (CNNs) to advance event detection, where only frame level static descriptors can be extracted by the existing CNN toolkits. This paper makes two contributions to the inference of CNN video representation. First, while average pooling and max pooling have long been the standard approaches to aggregating frame level static features, we show that performance can be significantly improved by taking advantage of an appropriate encoding method. Second, we propose using a set of latent concept descriptors as the frame descriptor, which enriches visual information while keeping it computationally affordable. The integration of the two contributions results in a new state-of-the-art performance in event detection over the largest video datasets. Compared to improved Dense Trajectories, which has been recognized as the best video representation for event detection, our new representation improves the Mean Average Precision (mAP) from 27.6% to 36.8% for the TRECVID MEDTest 14 dataset and from 34.0% to 44.6% for the TRECVID MEDTest 13 dataset.
报告人简介:Zhongwen Xu is a second-year Ph.D. student at Centre for Quantum Computation and Intelligent Systems (QCIS), University of Technology, Sydney (UTS), under the supervision of Dr. Yi Yang. Prior to that, he was with the He-Zhijun Honor Class in the College of Computer Science, Zhejiang University, and received his Bachelor’s degree from Zhejiang University in 2013. His research interests are computer vision and multimedia, particularly in deep learning for video analysis.

报告嘉宾2:常晓军(悉尼科技大学)
主持人:张利军(南京大学)
报告题目:Complex Event Detection using Semantic Saliency and Nearly-Isotonic SVM [Slides]
报告时间:2015年8月5日晚20:40(北京时间)
文章信息:
[1] X. Chang, Y. Yang, E. P. Xing, and Y. Yu. Proceedings of the 32nd International Conference on Machine Learning (ICML), 2015.
报告摘要: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.
报告人简介:Xiaojun Chang is a Ph.D student at Centre for Quantum Computation and Intelligent System (QCIS), University of Technology, Sydney, under the supervision of Dr. Yi Yang. His research interests are machine learning and its applications to multimedia and computer vision.

报告嘉宾3:闫岩(悉尼科技大学)
主持人:张利军(南京大学)
报告题目:Scalable Maximum Margin Matrix Factorization by Active Riemannian Subspace Search [Slides]
报告时间:2015年8月5日晚21:20(北京时间)
文章信息:
[1] Yan, Y., Tan, M., Tsang, I., Yang, Y., Zhang, C., and Shi, Q. Scalable Maximum Margin Matrix Factorization by Active Riemannian Subspace Search. International Joint Conference on Artificial Intelligence (IJCAI), 2015.
报告摘要:The rapid increase of Web services has witnessed an increasing demand for predicting the preferences of users on products of interest, such as movies and music tracks. This task is known as the collaborative filtering (CF) problem, which is a heated topic in recommendation systems. There are many ways to deal with the CF task. Based on the fact that users tend to share the same or similar preference over the products, low-rank property is then introduced to the problem and becomes very popular. Traditional nuclear-norm-minimization is effective, but not rather efficient due to the computationally expensive large-scale singular value decomposition (SVD). Hence, matrix factorization (MF) is proposed to recover the rating matrix by factorizing the original matrix into two small factor matrices, which is often scalable.
As for the user rating data, in general, they are given in the form of discrete values, including binary ratings and ordinal ratings The binary ratings can be either “+1” (like) or “-1” (dislike); while the ordinal ratings are in discrete values such as 1-5 “stars”, which are more popular in real world applications. To give more accurate prediction of such rating data, maximum margin matrix factorization (M3F) was proposed. This approach aims to find a margin between every two ordinal rating values, such as 3 and 4. Because of the ratings are in discrete values, M3F often achieve promising results. Existing M3F algorithms, however, either have massive computational cost or require expensive model selection procedures to determine the number of latent factors (i.e. the rank of the matrix to be recovered), making them less practical for large scale data sets. To address these two challenges, in this paper, we formulate M3F with a known number of latent factors as the Riemannian optimization problem on a fixed-rank matrix manifold and present a block-wise nonlinear Riemannian conjugate gradient method to solve it efficiently. As for the problem of detecting the number of latent factors, we then apply a simple and efficient active subspace search scheme to automatically estimate the rank of the matrix to be recovered. Empirical studies on both synthetic data sets and large real-world data sets demonstrate the superior efficiency and effectiveness of the proposed method.
报告人简介:Yan Yan received the B.E. degree in Computer Science and Technology from Tianjin University, Tianjin, China, in 2013. He is currently pursuing the Master degree with University of Technology, Sydney, Australia under the supervision of Dr. Yi Yang. His current research interest include computer vision and machine learning.

(Visited 770 times, 1 visits today)