Real-Time People Counting That Reduces the Load on Security Teams
Real-time people tracking and counting from CCTV video, for an AR and AI platform company
About the customer
Our client is a US technology start-up with an augmented reality (AR) and artificial intelligence (AI) platform for aerospace technicians, which speeds up defect documentation and validation. We built a people tracking and counting module for the platform. It uses CCTV video to monitor individuals entering and exiting a premises, reducing the workload on security teams.
What needed solving
Manual monitoring
Tracking people across large premises relied heavily on manual monitoring
Error-prone and resource-intensive
Manual methods were resource-intensive and prone to human error
Real-time need
The client needed automated tracking and counting of people in real time
How we built it
Detecting people
The YOLOv8 object detection model (from Ultralytics) identifies individuals in the video.
Tracking
The BoT-SORT tracker follows each identified individual, so entries and exits can be counted.
Video pipeline
A Python and Flask pipeline processes the video streams from CCTV sources.
Hosting
Paperspace virtual machines provide scalability and reliability.
Performance
CUDA and PyTorch speed up processing, with OpenCV and Scikit-Image for image processing.
Less burden on the security team
Technologies
Domains
Have a Query? Let's Have a Conversation
Do you have a project in mind? We’d love to know what it is. Email us or send a message from our contact page.
Have a similar problem?
Tell us what you are trying to solve and our team will get back to you.