Commercial
AI and ML-based solutions
R&D. Development. Consulting.
MENU
Multisense

Facial Anti-Spoofing with OpenCV

HOME / OUR WORK / Multisense
/ task
To improve the security of corporate devices even in remote environments. This technology offers a new approach to protect sensitive information by ensuring that only an authorized user can access their device using advanced facial recognition.
/ Solution
We implemented seamless Windows logon integration, on PCs of more than 1000 corporate users. Developed an anti-spoofing model with an error rate of less than 1% and a processing speed of less than 1 second.

This technology offers a new approach to protecting sensitive information and allows us to ensure that the right person is in front of the camera using advanced facial recognition.

Since the beginning of our project, 3 different face verification models have been developed: active, passive and flash model. All 3 models achieve accuracy above 99% and are optimized for all kinds of platforms with minimal latency in terms of processing time. Currently, work is ongoing to improve the accuracy of the models in complex environments and ambiguous perspectives, as well as to improve the API performance.
/ technologies
Python, PyTorch, Zephyr/Mistral LLMs, Docker, Git, Linux, OpenAI API, Flask, FastAPI