This is a addon for DOODS2, a separate service that detects objects in images. It's designed to be very easy to use, run as a container and available remotely.
A new field in the monitor settings will appear when the DOODS addon is enabled.
Enable for this monitor.
Individual confidence thresholds for each object that can be detected. A threshold of 100 means that it must be 100% confident about the object before a event is triggered. 50 is a good starting point.
Crop frame to focus the detector and increase accuracy.
Mask off areas you want the detector to ignore. The dark marked area will be ignored.
Objects with an percentage area smaller than this value will be discarded.
Objects with an percentage area greater than this value will be discarded. This filter will be disabled if the value is zero.
TensorFlow model used by DOODS to detect objects.
Frames per second to send to detector, decimals are allowed.
The number of seconds the recorder will be active for after a object is detected.
If sub stream should be used instead of the main stream. Only applicable if Sub input is set. Results in much better performance.
If you use Docker compose or bundle, then DOODS2 should already be running and all you should have to do is enable the addon. If you installed OS-NVR the bare-metal way, you need to install DOODS2 manually.
Start The DOODS2 service
sudo docker run -p 8080:8080 curid/doods2_tf-cctv:latest
Check if the service is working
curl 127.0.0.1:8080/version
Config file will be generated at configs/doods.json on first start after the addon has been enabled.