Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Engage Grants Program

University of Manitoba

$25,000.00

Object detection from security cameras for smart homes

FY 2016-17

Summary
Sengled, the partner company of the project, is a world leader in technologies and products for smart home. One of the most important products of the company is a smart light bulb combined with sound systems, security cameras and other smart features. It allow consumers to take control of the home's audio, security, connectivity simply by installing a light bulb.The security camera from the smart light bulb produces massive quantities of surveillance videos. However, the usefulness of these videos is limited by our ability to interpret and make decisions from them. The goal of this project is to develop computer vision technologies for detecting certain objects of interest (e.g. person, car, bicycle, cat, etc) from these surveillance videos. A reliable solution of this problem will enable the partner company to make full use of the surveillance videos they have collected. It will also help the company to develop more advanced video analytic products and services. For example, if we can reliably detect persons from the security cameras, the smart light bulb can automatically adjust the lighting conditions in different rooms. The surveillance videos from the company pose some unique challenges for standard object detection algorithms. First of all, these videos often have compression distortion, barrel distortion, night vision noise, etc. Second, these videos do not have large quantities of detailed object annotations. In the proposed research, we will study two specific problems. First, we will develop algorithms that can localize objects in weakly labeled videos, where the detailed object bounding box annotations are unavailable. Second, we will develop algorithms that can learn object models from multiple data sources. The proposed research will help the company to build reliable object detectors for its surveillance videos.
Program
Engage Grants Program
Location
Manitoba, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
500888-2016
Name nom
Wang, Yang
Department département
Computer Science
Organization id
19
Institution établissement
University of Manitoba
Province en
Manitoba
Province fr
Manitoba
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2016
Competition year année de concours
2016
Award amount
25000
Program id
EGP
Program name en
Engage Grants Program
Program name fr
Programme de subventions d'engagement partenariat
Group en
Research partnerships programs
Group fr
Programmes de partenariat
Committee code
1555
Committee name en
Prairie Internal Decision Committee
Committee name fr
Comité de décision interne pour les Prairies
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2800
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Artificial intelligence (computer vision, use2603)
Research subject fr
Intelligence artificielle (vision artificielle, utiliser 2603)
Partie
1
Nb partie
1
Application title
Object detection from security cameras for smart homes
Keyword
NSERC
Application summary
Sengled, the partner company of the project, is a world leader in technologies and products for smart home. One of the most important products of the company is a smart light bulb combined with sound systems, security cameras and other smart features. It allow consumers to take control of the home's audio, security, connectivity simply by installing a light bulb.The security camera from the smart light bulb produces massive quantities of surveillance videos. However, the usefulness of these videos is limited by our ability to interpret and make decisions from them. The goal of this project is to develop computer vision technologies for detecting certain objects of interest (e.g. person, car, bicycle, cat, etc) from these surveillance videos. A reliable solution of this problem will enable the partner company to make full use of the surveillance videos they have collected. It will also help the company to develop more advanced video analytic products and services. For example, if we can reliably detect persons from the security cameras, the smart light bulb can automatically adjust the lighting conditions in different rooms. The surveillance videos from the company pose some unique challenges for standard object detection algorithms. First of all, these videos often have compression distortion, barrel distortion, night vision noise, etc. Second, these videos do not have large quantities of detailed object annotations. In the proposed research, we will study two specific problems. First, we will develop algorithms that can localize objects in weakly labeled videos, where the detailed object bounding box annotations are unavailable. Second, we will develop algorithms that can learn object models from multiple data sources. The proposed research will help the company to build reliable object detectors for its surveillance videos.