With the amount and quality of video data collected increasing every day, Video Analytics (VA) algorithms are gaining popularity in use-cases that include situational awareness, post-incident investigation, real-time analysis and monitoring, and extraction of insights and trends for further analysis.
VA is already being used in counting number of pedestrians to assess pathway effectiveness, counting number of fish feed to optimise the operations of aquaculture hatcheries, and many others.
Government agencies who are new to VA often find themselves these questions: How can we kickstart a VA project quickly and effortlessly? How can we leverage our own and other agencies’ videos for analysis? Is it possible to develop our own VA models without being locked in by private commercial vendors? How can we make use of the existing VA algorithms by GovTech?
Having worked with multiple agencies on a myriad of CCTV and VA requirements, GovTech’s Data Science & Artificial Intelligence Division (DSAID) developed the Video Analytics System (VAS), a central platform for agencies to meet their VA needs.
What is VAS?
As a central platform, VAS aims to:
- Enable agencies to develop and deploy VA solutions.
- Exchange video data and analytics expertise across the whole-of-government (WOG).
- Expand the video sensor network and insights derived.
The key components of VAS are:
Media Hub (video sharing platform)
- Supports the upload, playback, search, request, and sharing of videos.
- Stream videos from cameras for on-demand viewing.
Insights Service (VA-as-a-service)
- Generate insights for your videos based on readily developed VA services for detection and counting of relevant events and objects
- Currently available VA Services:
- Smoking action detection
- Personal Mobility Devices (PMDs) / pedestrians / cyclists’ detection
- Vehicle Dwell Time analysis and number plate recognition
- Allows customisation of regions of interest.
- Generate shareable reports and be notified when the analysis is completed.
- Export reports in CSV format for custom analysis.
VA Enablement Service (VideoIO)
- Train VA model based on a selection of common deep learning model options to detect objects of interest.
- Curate and manage custom datasets for the training process.
- Review and perform annotation for both video and images.
- Initial and monitor model training runs and be notified when the training has completed.
- Test VA model by performing predictions using custom trained model versions.
- Export prediction results to CSV/JSON formats.
- Search and filter aggregated metadata according to available attributes, such as Vehicle Type (e.g., bicycle or Personal Mobility Device (PMD)), Event Type Attribute (e.g., smoking), and Spatial Attributes (e.g., postal code).
|Technical specifications of VAS|
|Hosting Platform||Government on Commercial Cloud (GCC)|
|Architecture Infrastructure Standard||GovTech Application Infrastructure Architecture Standard (AIAS) “H” Model|
|Instruction Manual||IM8 (ICT&SS Management)|
|Data Classification||Restricted and below|
|Data Sensitivity||Sensitive (Normal)|
|System requirements of VAS|
|Device||Government Standard Image Build (GSIB)|
|Minimum Video Resolution||4CIF – 704 x 576p|
|Minimum Framerate||6 FPS|
|Video Container Formats||MP4, MKV, FLV, MJPEG|
|Video Codec Formats||H.264, H.265|
|Image Formats||JPG, JPEG, PNG|
|Annotation Formats||PASCAL VOC|
Why Should VAS be Adopted?
The benefits provided by VAS include:
- Quicker enablement of VA tasks: Agencies with VA use-cases can benefit from the get-go with the accessible and intuitive system. Users appreciate the ability to quickly generate insights and analysis from videos collected during trials, POCs, and pre-analysis work.
- Reduction in manual analysis: With VAS’ automated analysis processes, agencies can unlock critical time and cost savings. For example, to help fish hatcheries optimise the usage of artemia (a type of microorganism used as fish feed), officers from the Singapore Food Agency (SFA) had previously reliedon manual counting of artemia, which can number around 1,000 per sample. With VAS, SFA officers now spend considerably less time conducting laborious manual counting.
- Designed for agencies, by agencies: Various Design Thinking workshops were held with agencies to seek their inputs. The system is also designed to respond quickly to changes. In sum, VAS is a product that is centered around what agencies need.
How Do You Use VAS?
Onboarding of the VAS is by invitation only. Interested agencies can fill up this form.
In response to the COVID-19 pandemic, Singapore implemented Circuit Breaker measures in 2020 to contain the spread of the virus. These measures included closure of non-essential workplaces and prohibition of dining-in at eateries. During this Circuit Breaker period, many Singaporeans sought respite in parks and green spaces. As a result, popular parks were congested.
In response, NParks and GovTech collaborated to develop Safe Distance @ Parks. It is a website that provides real-time information on crowd density at Singapore’s green spaces. The website allows users to check crowd density before visiting a park. It also allows the authorities to monitor and prevent the parks from being too crowded.
VAS is a crucial component of Safe Distance @ Parks. It can be understood as the link between physical hardware and the Safe Distance @ Parks website. Input from CCTV cameras physically installed in the parks are sent to VAS for analysis. Thereafter, the analysed information is fed to and displayed on Safe Distance @ Parks.
Work is in progress for VAS to support the following features in future releases:
Insights Service (VA-as-a-service)
- Marine vessel / kayak detection
- Safe distancing measurements and group detection
- Compliance checks for biometric photo enrolments
- Selected Commercial off the Shelf (COTS) solutions
- Enhanced insights reports and API offerings
For enquiries, fill in this form.
Meet the Team!
Christopher Sia, Project Director
Christopher has more than 15 years of industry experience ranging from software development, software testing, project management, business development, to strategic planning. He leads the Engineering team and oversees the development of VA products and solutions. VAS is one such product. Aside from driving product development, Christopher is also passionate about building meaningful partnerships and communities. He is also active in guiding young aspiring engineers who are interested in a career in the field of A.I. Christopher graduated with a Bachelor’s degree in Computer Engineering and a Master of Business Administration (MBA) from Nanyang Technological University.
Henry Lim, Lead Product Manager
Henry is a lead product manager in the VA team at GovTech, where he leads product development of the VAS platform to support the public sector in tapping into the potential of VA to better serve public needs. He has over 20 years of diverse industry experience across product management and technical development on global products across multinational organisations. Henry graduated with a degree in Computer Science from the National University of Singapore and is passionate about building products that makes technology accessible for improving everyday lives.
Lee Ning Sung, Lead Product Manager
Ning Sung is one of the lead product managers in the VA team at GovTech, where she supports Singapore's public sector in their adoption of VA to improve productivity. She has over 12 years of diverse experience in the Singapore public sector, ranging from fiscal strategy and organisation transformations, strategic planning and foreign diplomacy to product and artificial intelligence. She believes in driving productivity for the greater good. Ning Sung graduated with a BSc in Electrical and Computer Engineering and holds a double major in Human Computer Interaction from Carnegie Mellon University in Pittsburgh, USA, and she obtained her MEng in Electronics Engineering from the University of Tokyo, Japan.
Chua Teck Wee, Lead AI Engineer
Teck Wee is a Lead AI Engineer in the VA team under the Data Science and Artificial Intelligence Division (DSAID) at GovTech. He leads the development of custom VA solutions for agencies. Teck Wee graduated with a PhD in Electrical and Computer Engineering from the National University of Singapore. He enjoys building AI solutions for public good.
Ng Yong Kiat, Agency Engagement Lead
Yong Kiat is a team lead for engagements activities in on-boarding agencies onto WoG VA products and services. He graduated from the National University of Singapore with a MSc in Management of Technology and a Bachelor of Computing in Computer Engineering. He has over 17 years of experience in systems engineering, solutions architecting, and pre-sales. Yong Kiat is passionate about harnessing cutting edge technologies to bring value to various stakeholders.
Other VAS team members:
- Bevin Seetoh (VideoIO Full Stack Engineer)
- Bill Cai (DevOps Technical Lead)
- Chia Tewan (AI QA Engineer)
- Goh Kwang Huang (CVX Technical Lead)
- Henry Luo (VideoIO Technical Lead)
- Jessica Foo (AI QA Engineer)
- Muthiah Nachiappan (Insight Service Technical Lead)
- Muhammad Nazar (DevOps Engineer)
- Rachel Shong (Agency Engagement Manager)
- Richard Yan (DevOps Engineer)
- Ruth Cheng (Agency Engagement Manager)
- Tan Kai Wei (AI and Software Engineer)
- Tham Shi Yi (AI QA Engineer)
- Wang Yadong (Metadata Standards, QA Engineer)
Last updated 29 December 2021
Thanks for letting us know that this page is useful for you!
If you've got a moment, please tell us what we did right so that we can do more of it.
Thanks for letting us know that this page still needs work to be done.
If you've got a moment, please tell us how we can make this page better.