CASE STUDY | PRAIA GRANDE
PRAIA GRANDE AT-A-GLANCE
Location: Praia Grande, São Paulo, Brazil
Deployment Type: Public Safety
Use Cases: Surveillance, Watchlist Monitoring, Analytics
A smart city on the São Paulo coast, Praia Grande hosts more than 2 million tourists a year.
The city of Praia Grande has more than 2,500 cameras installed throughout public places, protecting roads, the beachfront, and public buildings. Like all tourist cities, Praia Grande also suffers from the presence of criminals and tourist-targeted theft, including pickpocketing. Despite the city’s database of wanted individuals and large number of cameras, manually monitoring for each person of interest was an arduous and inefficient task — not least because many use disguises to hide from authorities.
Mayor Alberto Mourão sought to reduce the crime rate so the beach town would be a more desirable destination for visitors and safer for all inhabitants. Face recognition was chosen as a practical solution for increasing security because of its ability to quickly and accurately identify persons of interest and suggest matches to surveillance staff for further review. With face recognition as a force multiplier, Praia Grande would have a better chance of detecting criminals throughout the city before they had a chance to cause any harm.
The SAFR Solution
Sandro Pardini, the deputy secretary of administration responsible for testing new security technologies, worked in partnership with Digifort to choose a reliable facial recognition software that was internationally recognized and tested by NIST, with an accuracy rate above 99.8% for all races. SAFR was selected due to its superior performance on live video and low bias rates as tested by NIST. Praia Grande deployed SAFR as a test security enhancement for the popular Estação Verão Show, a month-long music festival attracting daily crowds of 20,000 people.
During this test deployment, the SAFR system’s accuracy rate and speed exceeded expectations — all with a lower total cost of ownership than other facial recognition systems tested. SAFR added superior situational awareness throughout the event, easily detecting thousands of faces in the crowd, comparing them to a watchlist of registered criminals, and alerting surveillance staff to potential threats. If no match was found, no personally identifiable information (PII) was collected or stored, but an anonymous analysis of the crowd’s demographics was maintained.
After the test deployment, Pardini worked with Digifort to install SAFR in cameras across the city’s public squares, tunnels, and seafront. Since the implementation of monitoring, including face recognition, Praia Grande’s crime rate has fallen by more than 60% in monitored areas and crime in public buildings has been all but eliminated.
Effects of Monitoring
Reduction in crime across monitored areas
Accuracy with the Civil Police watchlist of registered criminals
Almost no criminal activity in public buildings
“Another fundamental aspect [of facial recognition] is the requirement of hardware, which can make the project more expensive. SAFR presents a giant advantage, since it requires only a quarter of the hardware its competitors require. This is especially significant in large-scale deployments.”
Digifort President Carlos Eduardo Bonilha
Choosing a Facial Recognition System That Actually Works
While many facial recognition systems can accurately match a visa photo to a well-lit, center-posed static image of the same person, they don’t perform well under the challenges of live video.
To guarantee the highest effective accuracy for live video, choose a computer vision solution with proven real-world performance:
Even with changes in appearance
(hairstyles, weight, aging)
<100ms Recognition Speed
Even with many faces on camera and large watchlists
People counting, age, gender, sentiment, attention, dwell time, and more
Deploy in the cloud, on premises, or embedded directly in a camera. Integrate with top VMS providers, existing hardware, and access control systems
How can SAFR help you respond to threats in real time? Contact our team for a demo.