September 21, 2021

SD Times Open-Source Project of the Week: CompreFace

SD Times Open-Source Project of the Week: CompreFace

SD Times Open-Source Project of the Week: CompreFace

In an effort to make facial recognition applications more available, Exadel has announced the open source release of CompreFace. CompreFace is a face recognition system that can be integrated as a standalone service or deployed through the cloud. 

“Facial recognition technology is being rapidly adopted across a variety of industries, including retail, travel, and hospitality. However, not everyone has the machine learning experience necessary to develop and deploy facial recognition efficiently and effectively,” said Lev Shur, president of Exadel Solutions. “Exadel designed CompreFace as a facial recognition solution that can be leveraged no matter what level of machine learning knowledge the end-user has.”

Key features include:

  • Ability to set up without previous or extensive machine learning knowledge
  • Use of deep neural networks for high accuracy
  • UI panel with roles for access control
  • Easy integration through its face recognition API

“CompreFace executes real-time facial recognition from streaming video and photos. The recognition process is based on a neural network that uses 512 features to classify parts of the face and then uses those classifications to identify each person. CompreFace needs at least two photos uploaded into the Face Collection to perform this facial feature analysis. The service can recognize faces even when they’re partially covered,” the Exadel AI team wrote in a post

The team also added that the software can be used to advance security measures. For instance: making a smart entrance security system, detecting unauthorized access, authenticating someone using a mobile device, finding a missing person in a public place, and increasing border security.

Other use cases can include tracking work and school attendance, speeding up check ins, greeting guests, developing targeted advertising, and integrating it into look alike apps.