What resolution sub-stream should I use for AI analysis

1

105437

BIT Beta Team
Jun 8, 2015
2,320
1,141
I've been testing the built-in AI using the main stream and it's causing CPU to peg in certain scenarios. I'd like to try using the sub stream to help with this. What resolution for the sub stream would still allow for good AI analysis for people, vehicles and animals? Thanks
 
Snapshots sent for analysis are resized automatically to 640x640 pixels (the standard default input size for YOLO object detection models) regardless of whether your camera source stream is 1080p or 4K.

Because of that, it will struggle the further away objects are from the camera.
 
  • Like
Reactions: 105437
wittaj said:
Snapshots sent for analysis are resized automatically to 640x640 pixels (the standard default input size for YOLO object detection models) regardless of whether your camera source stream is 1080p or 4K.

Because of that, it will struggle the further away objects are from the camera.
Click to expand...
So using the sub stream for AI that is set to 704x480 (D1) should be okay?
 
You must log in or register to reply here.
Top Bottom