In my test zoneminder 1.36.x user interface, the first camera is labeled as "Monitor - Front Door - HiRes (1)". The second camera is the same camera, just the low-res feed from it... "Monitor - Front Door (2)".
In objectconfig.ini for the ML aspect of it... are those entries supposed to be... "[monitor-1]" and "[monitor-2]" or some strings relevant to the names above?
I understand from other threads that when zm_detect.py is initiated by the server processes, it takes some sort of default settings for most variables... and those settings can be overridden as a default, or at each camera (monitor) from within objectconfig.ini. I think that's right. If I'm wrong, let me know.
I am also trying to test and understand at the command line level.
Code: Select all
root@270ccccf8ee0:/config/hook# sudo -u www-data /var/lib/zmeventnotification/bin/zm_detect.py -c /etc/zm/objectconfig.ini -e 2548 -m monitor-1
[ WARN:0] global /root/opencv/modules/dnn/src/dnn.cpp (1442) setUpNet DNN module was not built with CUDA backend; switching to CPU
[s] detected:person:99% car:83% --SPLIT--{"labels": ["person", "car", "car", "car"], "boxes": [[242, 202, 322, 426], [418, 277, 484, 309], [428, 266, 454, 280], [362, 272, 412, 284]], "frame_id": "snapshot", "confidences": [0.9923852682113647, 0.8343825936317444, 0.44217103719711304, 0.31473708152770996], "image_dimensions": {"original": [1080, 1920], "resized": [450, 800]}}
1. The "resized" image that shows the detected objects is very small compared to the size of the original. I found a "resize=800" in the objectconfig.ini. I changed it to a resize=no... no luck. Changed it back and added a resize=no to a monitor stanza with no luck, as well, but I'm not sure if I have that stanza correct anyway.
2. What is the correct string to put in the [monitor-x] header for the cameras... just the number? or some alpha string relevant to the friendly name in the GUI.
3. When I run zm_detect.py (see above for command and output), I found a switch to specify a monitor... -m. Is the correct syntax -m 1? or -m <some string relevant to the monitor name>? or -m monitor-1?
4. Where can I find a list of detectable object types (person, car, truck, dog, cat, llama, etc)?
5. I understand I can override that list of objects-to-detect and locate using the object_detection_pattern variable in objectconfig.ini... and I also assume I can override it at the specific camera (monitor) level.. maybe it'll start working when I get that [monitor-x] stanza configured correctly. I've copied the monitor-999 example, and tried to customize it by getting it to wait 5 seconds before analyzing, but not sure if I am doing it correctly or not (the #my driveway is from the original example, and I assume it is comment only, and not used programmatically; correct me if I am wrong about that).
Code: Select all
[monitor-1]
# my driveway
match_past_detections=no
wait=5
object_detection_pattern=(person|car|truck|dog|cat)
resize=no