Hi, I received some emails about the new docker image that was released with a (mostly) functioning machine learning system along with the ES.
I though I'd post a message so that everyone trying it out can follow along (I am the author of the ES+hook system):
1. When the ES (Event Server =/usr/bin/zmeventnotification.pl) sends a notification, you can choose to embed a picture in the notification (can't be http or self signed HTTPS). That URL setting is picture_url in /etc/zm/zmeventnotification.ini. Note that the default certificates generated by this docker image will not work if you want picture notifications. You need to replace them with proper certificates like those from Lets Encrypt.
2. If you want to enable "hook_pass_image_path" path in zmeventnotification.ini, you need to be running ZM 1.33 or above. This docker image (as of Sep 20, 2019) is for 1.32 - the last stable ZM release. In short, it won't work. What this option actually does is when the ES invokes the hook script (/usr/bin/detect.py), it lets the script know the file system path of the current event that is being passed for object detection. This path is retrieved by the ES using a feature only available in ZM 1.33+. When /usr/bin/detect.py runs & if it finds an object of interest in the image(s), then it will write an image file to that path called 'objdetect.jpg' which is basically the image where it found an object and a bounding box around the object of interest. This is then shown both in zmNinja and ZoneMinder as an image with boxes around the object. None of this will work for you in ZM 1.32. It will magically start working when this docker image is upgraded to the next version of ZM.
3. Next up, after detection is run, and it found something, the result is sent back to the ES. Lets say its "detected: car". This is prefixed to the existing motion notes of ZM in the DB. This feature works in 1.32 as well as is _not_ related to hook_pass_image_path at all.
4. If you are having issues, please read the documentation on ES and the hooks - including how to run it in manual mode and enabling debug logs
Yes, it's a little convoluted the way it is today. I could have made it simpler, but its what it is at the moment
I don't follow this forum at all - stopped by to clarify a few things. If you have questions specific to the machine learning stuff, you may want to ask in the ZoneMinder slack forum incase you are not able to resolve it here.
Links of interest:
Event Notification docs
Machine Learning Hooks
ZM Slack Channel Joining Link
