WeDevelop - Story Details

Success Story

CVSD

A vision-powered digital solution for Social Distancing

HealthComputer Vision

The CVSD system allows people to move freely through the installation with full adherence to COVID guidelines and restrictions. This virtual system is simple and fast, and it doesn’t even have to be installed — anyone can use it within their current camera network.

The Challenge

Unfortunately, facilities sometimes have to close — especially in this COVID-related new normal. Our challenge was to allow them to re-open quickly and effectively without impacting public and employee safety. Anyone that enters the facility must know that the space is secure. Meanwhile, the business must keep moving and avoid closure.

The Solution

We worked on a number of possible solutions to this challenge. Once we decided on the best candidate, we began improving it, adding functions and making it easier for all users. This ongoing development was finalized and delivered within the client’s tight time frame.

About Software Development Methodology

Due to COVID-19, this internal product was developed to control social distancing through computer visual terminals. The solution needed to handle frame processing of real-time video integrated within the existing camera network. We managed to achieve all of this with no extra setup or hardware. The solution we built can be up and running by itself in 15 minutes.

This tool was able to carry out immediate statistical analysis of the extracted data, allowing the re-opening of facilities and providing a healthy space for the guests. We achieved this by tracking occupancy levels and the average distance between people.

We decided to gain customers’ trust by showing them this information in real-time. By printing a QR code and displaying this on the wall, we were able to deliver information about the room to visitors without a moment's delay.

Teams using the solution can access actionable metrics and get instant alerts based on activity in the facility.

Tech Stack

Deployment in a serverless automated environment.
Real time video frame processing.
Statistical analysis of the extracted data.
Visualization of video streams with information overlay.
Deployed in an environment severless (App Engine) automated.
Deployed manual of NVIDIA Triton (Al server) to solve the problems with the Google Cloud Al Platform.
Development of AI models, combining pre-existing ones with TensorFlow, as well as manual optimization of those models.
Google Cloud for all (App Engine, Al Platform, Compute Engine, DNS, BD, PubSub, Storage, Cloud Build).
Google Analytics.