CodeProjectAI is tagging items twice... Is that wasting resources?
- Thread starter erkme73
- Start date
@erkme73 Yes, CP.AI sent your image to multiple models to detect objects, your last screenshot list each model used at every timestamp and how long each prediction took. IMHO, those prediction times are on the slower side for my taste.
You can reduce resources by turning off default models and specifying 1 custom model in "Custom models" like:
You can also speed up CP.AI by turning off "Use main stream if available".
Turn off "Static object analysis" to reduce CP.AI load. Unless you regularly have parked cars in the above camera POV, you don't need it.
You can reduce resources by turning off default models and specifying 1 custom model in "Custom models" like:
objects:0,alpr:0,faces:0,ipcam-generalYou can also speed up CP.AI by turning off "Use main stream if available".
Turn off "Static object analysis" to reduce CP.AI load. Unless you regularly have parked cars in the above camera POV, you don't need it.
Last edited:
Excellent, thank you! I added the custom model to the camera's alert/ai setting, and disabled the custom under global. The CodeProjectAI dashboard is very quiet now, despite constant movement on the camera (lots of rain).
Is this along the lines of what you were suggesting?
Interestingly, I can get to the CodeProjectAI dashboard at the IP:5000 and IP:32168. I have to assume there is only one copy of the server running as the dashboard for both look identical on the status page/log.
I wasn't clear on how to disable the default models, or where to add the custom model (global or per-camera). I'm unable to change the global custom camera field as it is greyed out. And the "stop now" does nothing.
Is this along the lines of what you were suggesting?
Interestingly, I can get to the CodeProjectAI dashboard at the IP:5000 and IP:32168. I have to assume there is only one copy of the server running as the dashboard for both look identical on the status page/log.
I wasn't clear on how to disable the default models, or where to add the custom model (global or per-camera). I'm unable to change the global custom camera field as it is greyed out. And the "stop now" does nothing.
Great, thanks. Ok, turned on custom models under global, and changed port to 32168. Aside from that, looks good? I appreciate your help more than you know. I'm actually helping a neighbor with improving his AI alerts, so the more I learn here, the better for both of us.