CyberLink Logo

CyberLink Support

Back

What recognition models are recommended for different usage scenarios?

Face Extraction Models in FaceMe SDK:

  1. Ultra High Precision Model (UH):

    • Features: Requires NVIDIA GPU or Movidius, providing the highest accuracy.
    • Recommended For: Servers, workstations, or high-performance personal computers.
  2. Very High Precision Model (VH):

    • Features: Computational demand is only one-tenth of the UH model but with slightly lower accuracy.
    • Recommended For: Advanced mobile devices (Snapdragon 600 or higher) and common personal computers (Celeron or higher).
  3. High Precision Model (H):

    • Features: Designed for mobile devices and streamlined users, suitable for face front verification scenarios.
    • Recommended For: Mobile devices and low-power systems.

If your platform or usage scenario is not covered by the above models, please contact us for further information.

Find More Resources

Was this page helpful?