[Presentation] OpenVINO for BI

mpl

n3wb
Nov 6, 2023
25
13
Germany
Hello everyone, let me introduce: OpenVINO for Blue Iris.
Over the past few days I’ve been working with OpenVINO + YOLOv11 and put together my first working object and face detection setup.


Everything is running on the Intel NPU (Intel Core Ultra 5 245).


Current status:
  • Object detection
  • Face detection
  • Face recognition
  • Unknown faces are saved and currently need to be assigned manually:
    1. Move the image into the matching folder (person’s name)
    2. Run the learning/update script afterwards
      → this improves the recognition rate.

Whats for me still missing is a proper API integration with Blue Iris so the whole thing can be integrated more cleanly. Right now, the settings are still configured via a Python config.

Does anyone have the BI(6) API details or example requests for that?


Performance (NPU on my system):
  • YOLO nano: ~15%
  • YOLO small: ~20–25%
  • YOLO medium: ~25–30%
  • YOLO large: >30%

Do people actually need “large”, or is that overkill for everyday use?
Without face recognition it runs a little bit faster.

Would anyone be interested in this (e.g., code, setup guide, or a small release), or should I keep it as a private project?


PS: I also have a Coral USB Accelerator and a Dual Coral PCIe Accelerator incl. PCIe adapter for sale
 

Attachments

  • 1766268619546.png
    1766268619546.png
    1.8 MB · Views: 7
  • 1766269768954.png
    1766269768954.png
    1.4 MB · Views: 7
  • 1766271407219.png
    1766271407219.png
    1.5 MB · Views: 7