Human Attribute Extraction enables computers to recognizes human's gender,
age, and race as well as a few categories of apparels from images or
surveillance videos, which greatly facilitates automatic customer profile
analysis. These technologies have been commercialized by NES Soft, Ltd., and
integrated into the system Field Analyst 2.0, and NEC Digital Signage.
Detection and localization of events involving some specific human actions
from videos shows great potential to help elderly care, security monitoring,
and customer behavior analysis. Our event detection system was among the top
performers in the TRECVid evaluation for event detection 2008 and 2009.
Human Counting in Videos
Automatic human counting from video sequences provides valuable information
to surveillance and retail business. Our technology can accurately detect,
track and count humans walking in a variety of directions without having to
rely on frontal face detection. The technology have been integrated into two
NEC products: FieldAnalyst and Kaoato Surveillance Software by NEC Soft Ltd.
SmartCatch
SmartCatch is an intelligent video surveillance system invented at our
department. SmartCatch provides comprehensive video surveillance solution
with functionalities such as human detection and tracking, unattended object
detection and behavior recognition. The core technologies were successful
transferred into a spin-off company, Vidient Systems, in 2003, which has
become a global leader in sophisticated video analysis.
ViRMoS: Driving Danger Level Prediction
Internet OLAP (iOLAP)
Internet search engines help users to retrieve
their desired information through keyword-based search. However,
information retrieval covers only a small portion of the entire
spectrum of data access needs from all users. Quite often, instead
of retrieving a small subset of relevant data, users want to know
the overall picture, or want to visualize the distribution of the
entire data set. This kind of ability is more important for business
intelligence and decision making supports.
We are developing the Internet OLAP
technologies that aim to sift through millions of websites and blogs
to recognize/visualize the overall picture of the underlying
internet data set, and to discover value-added information such as
the constituent communities, the hot topics and their evolutions,
the influential people, etc. Such technologies will enable us to
leap beyond the keyword-based search to a deeper understanding of
semantics and structures of internet data sets, and will have huge
potential for grassroots knowledge acquisition, sentiment discovery,
market analysis, business strategy planning, etc.
Recommendation Systems
Social Network Analysis
Document Summarization
Quickly get the gist of a document, webpage, reviews, or any text
selection of your choice. Identify the key topics and highlight the
difference of documents while eliminating redundant information.
[ demo for product review |
demo for webpage summary
]
