How Casinos Catch Eight Times More Unwanted Visitors with Face Recognition.
No developer would argue, face recognition in crowd scenes from security cameras is a challenging task for several reasons:
- Low image quality; security cameras typically produce low-resolution images, primarily in poorly lit conditions, which can make it challenging for facial recognition algorithms to accurately identify faces;
- Occlusions. Faces in crowd scenes are often partially or completely obstructed by other people, objects, or accessories. This can further degrade the performance of facial recognition algorithms.
- Pose variation: Faces in crowd scenes can be at different angles and poses, which can make it difficult for facial recognition algorithms to match them to known faces.
- Lighting variation. Lighting conditions in crowd scenes can vary widely, which can make it difficult for facial recognition algorithms to accurately identify faces.
- Background clutter can also cause issues, as crowd scenes can be very cluttered, with many different people and objects in the background, making it difficult for facial recognition algorithms to isolate and identify faces.
But still, in casinos, where large sums of money are involved, the task of accurate identification of unwanted visitors without generating a plethora of false alarms is vital. Here's how one solution achieved this remarkable feat.
Omnigo, a Canadian company, offers one of the world's leading software solutions for public safety, incident reporting, and security management. Their product, "Unwanted Visitor," provides facial recognition from the casino's blacklist and is installed in over 350 casinos, including most in Las Vegas. However, they faced a challenge. They were on the lookout for opportunities to improve recognition accuracy rates and reduce false positives to enhance the efficiency of the Unwanted Visitor service.
In early 2019, Omnigo started testing alternative facial recognition algorithms, including one from 3DiVi. Preliminary tests on a database of 4,000 people and 700,000 test images showed promising results. This led to a pilot stage where the face recognition SDK was integrated into the security system of one of their casino clients.
The Results:
The test involved 1.05 million images of individuals entering two casinos over two weeks. The outcomes were then compared with the performance of their previous vendor. The past vendor identified 25 people from a hotlist of 21,476 individuals and dismissed 6,255 alerts. In contrast, the 3DiVi algorithm recognized 221 people from the same hotlist and rejected only 1,831 alerts. This implies that the 3DiVi algorithm identified nearly 8 times as many people with 70% fewer false alarms.
Yes, achieving these results would have been impossible without the dependable 3DiVi face recognition technology. However, the swift and responsive collaboration with the development team and the capability to precisely adjust the technology to suit the specific case also played a crucial role in its success.
This reaffirms that face recognition is not solely technology-driven, but is intertwined with business and relationships.
So if you're searching for a solid business partner and advanced face recognition technology, learn more about 3DiVi Face SDK, Face Recognition API and OMNI Platform.
✨ And of course, don't forget to experience 3DiVi Inc technology firsthand by giving it a try! Download the 3DiVi Demo App for your Android or iOS smartphone. You can find both versions ready for download.
✨ And of course, don't forget to experience 3DiVi Inc technology firsthand by giving it a try! Download the 3DiVi Demo App for your Android or iOS smartphone. You can find both versions ready for download.
