mikelud / codeproject.ai-custom-ipcam-models Goto Github PK

Hi,

Your model works really well for plate recognition day and night and OCR in BlueIris, would it be possible to incorporate your license plate model in Frigate NVR:

https://docs.frigate.video/configuration/advanced/#model

https://github.com/blakeblackshear/frigate/discussions?discussions_q=is%3Aopen++custom

https://coral.ai/docs/edgetpu/retrain-detection/

As the title says. Maybe a video guide?
Tried making one for my coral, but BI doesn't recognize it, and I'm seeing .yaml files for IPcam models. Not sure what I should do when making those.

I just switched to using your IPcam General and IPcam Animal models that come with Codeproject.AI. While they are much faster on my system than the default ones, I'm seeing a number of odd false positives. A water puddle is detected as 3 squirrels repeatedly (70%+ confidence), a gap between my shop and garage at night is detected as a person (60-70% confidence), my wife is detected as both a dog and a person simultaneously (80% and 91%), and there's a stray cat that's been detected as a cat, dog, person, fox, and bird (all with 70-90% confidence).

I'd be more than happy to send you some exported clips and Blue Iris .dat files where it had false positives if that would help improve their accuracy.

Does IPcam-combined include the dark images as well? I am unsure.

Do you have any models that can be used with the new CodeProject.AI Coral TPU module?

How do you install the Custom Models for TPU? The Coral Model doesn't have a custom model folder and I don't think it uses YOLO as far as I'm aware at least.

Hi Mike,

Appreciate your work very much.
Maybe an idea to make a model for detecting a heron? This would be very beneficial for people with a pond.

kind regards,
Dwayne