hardware based object detection.
New
We need some type of solution that add hardware based object detection to our servers.
integration with something like Codeproject.AI like how Blue Iris has implemented would be ideal. like this https://youtu.be/jqc1ukrkMmI?si=Y07gVeNpLxH58ADe&t=657
TPU/Detection hardware is becoming readily available.
OpenVino is nice, but its implementation is VERY limited. Vehicle, animal, Licence Plate recognition (with events based on pre determined licence plates) is stuff that would set NX apart.
Also detection hardware is getting REALLY cheap.
-
Yes I agree, to me its great that cameras can do analytics built-in and Nx Witness has plugins to talk to the camera, but what would be even better is let us put GPUs in the server and turn every single camera into smart cameras. With area intrusion (and of course other alarm types) drawn right in the software natively instead of the other brand's camera's software or in the little plugin window.
Even if we have to buy an "AI Professional License" per camera or something I think that is fair as long as it keeps improving with updates.
If this is already an existing feature or plugin maybe I am just out of the loop 😂
-
+1 vote for this feature, it would be a great benefit to many users such as myself without cameras that support analytics. It would be great not having to replace a whole fleet of cameras to support analytics if NX can do itself instead. I would also be open to paying a license cost to enable this feature per camera.
-
Hi
Have you had a chance to check the new Nx Toolkit?
There is a whole new development tool called Nx AI Manager.
It is probably fit your request and desire.
https://www.networkoptix.com/blog/2024/3/21/nx-ai-manager-edge-ai-pipelines-made-easy-and-instantly-scalable -
Another example of how this might work is how the open source Frigate NVR system does object detection with the Google Coral M.2 or USB accelerator keys. Those dedicated chips provide a lot more AI processing ability than a regular CPU. And cheaper than most GPU options.
Please sign in to leave a comment.
Comments
4 comments