[신간의 별자리x] 우리/미술, 그리고 ‘슬픔의 박물관’

Our goal is to develop a visual monitoring system that passively observes moving objects in a site and learns patterns of activity from those observations. For extended sites, the system will require multiple cameras. Thus, key elements of the system are motion tracking, camera coordination, activity classification, and event detection. In this paper, we focus on motion tracking and show how one can use observed motion to learn patterns of activity in a site. Motion segmentation is based on an adaptive background subtraction method that models each pixel as a mixture of Gaussians and uses an online approximation to update the model. The Gaussian distributions are then evaluated to determine which are most likely to result from a background process. This yields a stable, real-time outdoor tracker that reliably deals with lighting changes, repetitive motions from clutter, and long-term scene changes. While a tracking system is unaware of the identity of any object it tracks, the identity remains the same for the entire tracking sequence. Our system leverages this information by accumulating joint co-occurrences of the representations within a sequence. These joint co-occurrence statistics are then used to create a hierarchical binary-tree classification of the representations. This method is useful for classifying sequences, as well as individual instances of activities in a site.

/pdf/learning-patterns-of-activity-using-real-time-tracking-4pcdy8vvvh.pdf

Learning patterns of activity using real-time tracking

Most state-of-the-art techniques for multi-class image segmentation and labeling use conditional random fields defined over pixels or image regions. While region-level models often feature dense pairwise connectivity, pixel-level models are considerably larger and have only permitted sparse graph structures. In this paper, we consider fully connected CRF models defined on the complete set of pixels in an image. The resulting graphs have billions of edges, making traditional inference algorithms impractical. Our main contribution is a highly efficient approximate inference algorithm for fully connected CRF models in which the pairwise edge potentials are defined by a linear combination of Gaussian kernels. Our experiments demonstrate that dense connectivity at the pixel level substantially improves segmentation and labeling accuracy.

https://graphics.stanford.edu/projects/densecrf/densecrf.pdf

Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials

The PASCAL Visual Object Classes Challenge

An overview of the self-organizing map algorithm, on which the papers in this issue are based, is presented in this article.

The Self-Organizing Map

Under the three-year Video Surveillance and Monitoring (VSAM) project (1997‐1999), the Robotics Institute at Carnegie Mellon University (CMU) and the Sarnoff Corporation developed a system for autonomous Video Surveillance and Monitoring. The technical approach uses multiple, cooperative video sensors to provide continuous coverage of people and vehicles in a cluttered environment. This final report presents an overview of the system, and of the technical accomplishments that have been achieved.

http://www.ri.cmu.edu/pub_files/pub1/collins_robert_1999_1/collins_robert_1999_1.pdf

A System for Video Surveillance and Monitoring

http://haselab.info/soinn/img/pdf/03_NN-Shen1.pdf

An incremental network for on-line unsupervised classification and topology learning

https://cs.nju.edu.cn/rinc/download/2007_1.pdf

An enhanced self-organizing incremental neural network for online unsupervised learning

The paper presents a new online incremental zero-shot learning method for applications in robotics and mobile communications where attribute labeling is obtained via online interaction with users, and where the potential for inconsistency exists. Unique to most previous offline batch learning methods, the proposed method is based on the indirect-attribute-prediction (IAP) model instead of the direct-attribute-prediction (DAP). Using self-organizing and incremental neural networks (SOINN) as the learning mechanism, our method can learn new attributes and update existing attributes in an online incremental manner while retaining as high accuracy as that of the state-of-the-art offline method. Compared to the offline methods, the computation time has also been reduced by more than 99%. Two experiments evaluated two aspects of the proposed method. First, our method clearly outperforms the previous IAP-based offline method in terms of both time and accuracy, and yield approximately the same accuracy as the DAP-based offline method. Second, the proposed method can deal with situations where object attributes are gradually labeled via interaction with many users and where some of them may be incorrect. This scenario is very important for applications in mobile communications and robotics where some objects and attributes may be initially unknown and must be learnt online.

Osamu Hasegawa

Papers

A System for Video Surveillance and Monitoring

An incremental network for on-line unsupervised classification and topology learning

An enhanced self-organizing incremental neural network for online unsupervised learning

Online incremental attribute-based zero-shot learning

A fast nearest neighbor classifier based on self-organizing incremental neural network