AI-powered video-based real-time occupancy tracking on AWS

project image
Learn how we have developed the AI-based system that helped DTiQ customers (including retailers, restaurants, and gyms) during the COVID-19 outbreak to monitor location occupancy in real-time by leveraging the existing CCTV infrastructure.

Executive Summary

Learn how we have developed the AI-based system that helped DTiQ customers during the COVID-19 outbreak to monitor location occupancy by leveraging the existing CCTV infrastructure.

Key Figures:

  • Initial scope for the minimum viable product was delivered in 15 man-days.
  • Further phases of the project were delivered in time and budget.
  • Pattern Match contributed to the project in the cloud architecture area and significant parts of the software development.
  • Project is deployed live, and it is actively developed and provides revenue to the company.

About the Subject

DTiQ offers holistic solutions for the whole value chain of loss prevention and intelligent video solutions for retailers: smart auditing, fraud detection, loss prevention, customer engagement, preventive safety, and advanced analytics.

For this particular project you can learn about its objective and how it works from a following video explainer:

Challenges and Objectives

DTiQ Occupancy Monitoring service works by using a current location’s equipment, only by using the server to capture static images from the existing cameras at pre-defined intervals, sending them to the cloud where we analyze how many people are in the images. Our solution works for existing CCTV installations, for both IP-based and USB cameras.

DTiQ’s smart algorithm is looking for outlines of heads. Thanks to the method of using static pictures, it doesn’t matter if the person is standing, sitting, running, or eating. The whole solution is resilient even for the intersecting views from multiple cameras. It can be managed with the use of visibility zones that exclude tracking from a specific section for a given camera.

Once we have the total count, we compare against the pre-defined max and give you a percentage range of fullness. This can be displayed on an external panel by the doors or inside restaurant. Additionally it can be used in back-office applications.

To inform customers, DTiQ installs one of 4 options in the store. This can be outside the front door or inside, depending on the client’s needs. The digital sign will tell the customer if they can or cannot enter the restaurant.

The starting point of our cooperation was to evaluate the existing architecture, redesign the solution, and determine the next required steps. Right after the short design and research phase, Pattern Match joined the software development team that developed the final solution and actively participated in the future evolution and maintenance.

Some of the noteworthy achievements were:

  • redesigning the proposed architecture to serverless-based one to achieve lower operational cost,
  • deploying a secure and publicly available REST API for the partners and internal components with the use of Amazon API Gateway service,
  • improved monitoring - especially with additional, business-based metrics that gave a better insight into a system for non-technical team members,
  • scalable automation for the integration between customized DVR and Amazon Kinesis Video Streams,
  • development with cost-efficiency in mind from the beginning of the project,
  • introducing infrastructure automation with the use of AWS Serverless Application Model, AWS CloudFormation, and Jenkins-based CI/CD pipelines,
  • optimizing, deploying, and scaling existing machine-learning models with the use of Amazon SageMaker,
  • successful software development throughout the project life cycle,
  • introducing best practices for remote-first collaboration.

Benefits

Pattern Match cloud architects proved themselves as reliable and proactive partners with profound experience in AWS, Serverless, Machine Learning, and Serverless. Additionally, as our engineers have broad experience in various technologies and domains, we were able to help our clients with cloud-related challenges and bespoke software development for Java and Python.

Pattern Match, from the early beginning, showed their value with our transformation to serverless architecture. Vast knowledge about AWS that Pattern-Match has allowed us to avoid many common problems for the newbies in this technology and infrastructure. They support our team on each stage of our project, starting from planning, architecture, and finish on implementation. We are glad to have such a partner, and we recommend them as professionals in Serverless Computing and AWS.

Krzysztof Osiński

VP of Research and Development at DTiQ

Our cloud architects are responsible for the design, development, and maintenance of the platform. Depending on the period, Pattern Match contributed to about 40-75% of the workforce of the project. Thus, it was crucial to maintain close communication and active cooperation, both remotely and with client’s Gliwice-based employees.

Results

The project was delivered in time and budget. After 3 weeks of development, we have started to onboard clients to the platform on subscription-based commercial agreements. Over time, we have improved the performance, cost-efficiency, and scalability of the whole solution. The first phases of the entire project were delivered successfully. What’s also relevant, is that Pattern Match helped to deliver the product with the new technologies, methodologies, built-in security, and best practices. Everything in collaboration with the team, where such an experience gathered, will be beneficial to all future initiatives.

Want to hear more?

CONTACT US