Recognizing people in images is one of the foremost challenges in computervision. It is important to remember that consumer photography has a highly social aspect. The photographer captures images not in a random fas...
详细信息
Recognizing people in images is one of the foremost challenges in computervision. It is important to remember that consumer photography has a highly social aspect. The photographer captures images not in a random fashion, but rather to remember or document meaningful events in her life. The culture of the society of which the photographer is a part provides a strong context for recognizing the content of the captured images. We demonstrate one aspect of this cultural context by recognizing people from first names. The distribution of first names chosen for newborn babies evolves with time and is gender-specific. As a result, a first name provides a strong prior for describing the individual. Specifically, we use the U.S. Social Security Administration baby name database to learn priors for gender and age for 6693 first names. Most face recognition methods do not even consider the name of the individual of interest, or the name is treated merely as an identifier that provides no information about appearance. In contrast, we combine image-based gender and age classifiers with the cultural context information provided by first names to recognize people with no labeled examples. Our model uses image-based age and gender estimates for assigning first names to people and in turn, the age and gender estimates are improved.
This paper presents a general framework for live video analysis. The activities of surveillance subjects are described using a spatio-temporal vocabulary learned from recurrent motion patterns. The repetitive nature o...
详细信息
In this paper we present parallel implementations of some representative low level vision algorithms on a cluster of workstations. These include convolution operation and the image restoration algorithm using Markov r...
详细信息
Tracking humans in an indoor environment is an essential part of surveillance systems. vision based and mirophone array based trackers have been extensively researched in the past. Audio-visual tracking frameworks hav...
详细信息
The proceedings contain 525 papers. The topics discussed include: learning visual similarity measures for comparing never seen objects;a contextual dissimilarity measure for accurate and efficient image search;learnin...
详细信息
ISBN:
(纸本)1424411807
The proceedings contain 525 papers. The topics discussed include: learning visual similarity measures for comparing never seen objects;a contextual dissimilarity measure for accurate and efficient image search;learning local image descriptors;principal curvature-based region detector for object recognition;a benchmark for the comparison of 3-D motion segmentation algorithms;a nonparametric treatment for location/segmentation based visual tracking;learning gaussian conditional random fields for low-level vision;hierarchical structuring of data on manifolds;learning GMRF structures for spatial priors;element rearrangement for tensor-based subspace learning;unsupervised clustering using multi-resolution perceptual grouping;object tracking by asymmetric kernel mean shift with automatic scale and orientation selection;and free-form nonrigid image registration using generalized elastic nets.
We propose the creation of an e-research facility with the working title European laboratory for software evolution (ELSE). Our motivation is that a significant obstacle to progress in researching evolutionary phenome...
详细信息
We propose the creation of an e-research facility with the working title European laboratory for software evolution (ELSE). Our motivation is that a significant obstacle to progress in researching evolutionary phenomena in software-intensive systems is the difficulty of conducting realistic, industrial-scale investigations, especially for small research teams operating at armpsilas-length from each other. ELSE will address this problem by providing a virtual research environment (VRE) where researchers in software evolution and related fields can find annotated collections of data, tools, patterns and templates, dasiahow-topsila documents, previous results, and access to expertise in using them. The Laboratory will provide an on-line environment where researchers and practitioners can share research facilities and experiences and seek partners in projects. ELSEpsilas concerns will encompass both the social and the engineering aspects of evolution in software-intensive systems, and will therefore draw on existing achievements in both e-social science and e-science. It will create an exciting and innovative facility that will be attractive to researchers in fields ranging from software engineering, through informatics and business history to society and technology studies and anthropology. We are actively seeking both academic and industrial partners who will join us in taking this vision forward.
In this paper we apply approximate and multiscale entropy metrics to spectrum occupancy data gathered during an extensive measurement campaign. We show that the presented methods can be successfully applied to search ...
详细信息
In this paper we apply approximate and multiscale entropy metrics to spectrum occupancy data gathered during an extensive measurement campaign. We show that the presented methods can be successfully applied to search for and quantify structures of highly varying complexities and time scales. Although structures can be found they are unfortunately relatively complex and it does not appear to be easy to directly exploit them for dynamic spectrum access. We also discuss the major foreseen application areas for entropy-based analysis. These include increasing the reliability of spectrum sensing as well as validation of spectrum occupancy models. Finally, we highlight some results that should be taken into account in future work on spectrum sensing. In particular we observe that strong interference has significant structure, and cannot be well approximated by noise.
Current methods for learning visual categories work well when a large amount of labeled data is available, but can run into severe difficulties when the number of labeled examples is small. When labeled data is scarce...
详细信息
ISBN:
(纸本)9781424411795
Current methods for learning visual categories work well when a large amount of labeled data is available, but can run into severe difficulties when the number of labeled examples is small. When labeled data is scarce it may be beneficial to use unlabeled data to learn an image representation that is low-dimensional, but nevertheless captures the information required to discriminate between image categories. This paper describes a method for learning representations from large quantities of unlabeled images which have associated captions;the goal is to improve learning in future image classification problems. Experiments show that our method significantly outperforms (1) a fully-supervised baseline model, (2) a model that ignores the captions and learns a visual representation by performing PCA on the unlabeled images alone and (3) a model that uses the output of word classifiers trained using captions and unlabeled data. Our current work concentrates on captions as the source of meta-data, but more generally other types of meta-data could be used.
暂无评论