Stepping through zmeventnotification setup

Support for the docker image maintained by dlandon
Post Reply
tlundste
Posts: 19
Joined: Sun Nov 15, 2020 3:41 am

Stepping through zmeventnotification setup

Post by tlundste »

I'm trying to go through the setup to enable better notification via the YOLO, FACE, options, etc... I'm tired of partly-cloudy windy days running shadows fast across my zones and setting off event detection. And, I want to play with the "person" "dog" "car" etc object detection.
First of all... after I enabled the "=1" settings of my ZM image/container, it takes nearly an hour to start the image and container. Most of it is waiting on "compiling opencv without GPU support" (that step alone is around 40 minutes!). But, it comes up.

First question... is there something I can do to speed up the start-up of the container?
Second question... What am I missing to open up "https://<myIP>:9000/zm"? I just get "this page is not working". If I go to :9080, the apache webserver main page comes up. If I go to ":9443", I get my main Zoneminder interface. But nothing at port 9000.
Initially, when I brought up the container after setting the values in the container for YOLO, etc to =1, port 9000 was listening in my Ubuntu host, but when I "exec'd" into the container, port 9000 was not listening, nor was zmeventnotification.pl running. I did some digging, enabled opt_use_eventnotification in ZoneMinder's options via the web interface, and restarted the image and container. This time, I could see port 9000 listening and the .pl file running inside the container. Yet, I still can't get to that :9000 webpage.

Here is my config...
Ubuntu LTS 20.04.1 LTS
Hardware is a NUC8i3
Running docker, with docker-compose setup.
Zoneminder functions just fine. Here are the relevant docker-compose.yaml sections.

Code: Select all

############# ZONE MINDER

 zm_db:
   container_name: zm_db
   restart: always
   image: mysql/mysql-server:latest
   environment:
     - TZ=America/Chicago
     - MYSQL_USER=zmuser
     - MYSQL_PASSWORD=zmpass
     - MYSQL_DATABASE=zm
     - MYSQL_ROOT_PASSWORD=mysqlpsswd
     - MYSQL_ROOT_HOST=%
   network_mode: znet

 zoneminder:
   container_name: zoneminder
   image: dlandon/zoneminder:latest
   restart: unless-stopped
   depends_on:
     - zm_db
   shm_size: 4096m
   ports:
     - 9443:443/tcp
     - 9000:9000/tcp
     - 9080:80
   network_mode: znet
   environment:
     - TZ=America/Chicago
     - ZM_DB_HOST=zm_db
     - SHMEM=50%
     - PUID=99
     - PGID=100
     - INSTALL_HOOK=1
     - INSTALL_FACE=1
     - INSTALL_TINY_YOLOV3=0
     - INSTALL_YOLOV3=0
     - INSTALL_TINY_YOLOV4=1
     - INSTALL_YOLOV4=0
     - MULTI_PORT_START=0
     - MULTI_PORT_END=0
   volumes:
     - /mnt/server/zoneminder:/config:rw
     - /mnt/server/zoneminder/data:/var/cache/zoneminder:rw
I've done nothing with the supplied .ini files, yet... Should I be able to get to the :9000 page without modifying any of the .ini files, first? If not, what needs modifying?
Magic919
Posts: 1381
Joined: Wed Sep 18, 2013 6:56 am

Re: Stepping through zmeventnotification setup

Post by Magic919 »

Port 9000 is not a web page. It's for communication with ZMES.
-
tlundste
Posts: 19
Joined: Sun Nov 15, 2020 3:41 am

Re: Stepping through zmeventnotification setup

Post by tlundste »

OK... thank you. I'll keep reading to see if I can figure out what's next, then. I thought I saw someone post that :9000 as a means to access something within, via a web browser.
Still looking to find out if there is a way to start up my container more quickly. Before enabling the "YOLO" and other settings, it would start in a few minutes... now, it's nearly an hour.
Magic919
Posts: 1381
Joined: Wed Sep 18, 2013 6:56 am

Re: Stepping through zmeventnotification setup

Post by Magic919 »

If part of the startup is that it downloads the source, configures and then builds Opencv, then it'll take a while. If the build is not optimised to use all the available cores and/or the machine lacks grunt it'll only be worse.

If the next steps involve it installing Yolo and other ZMES goodness, then more of the above. This bit is much smaller impact than the Opencv step.
-
tlundste
Posts: 19
Joined: Sun Nov 15, 2020 3:41 am

Re: Stepping through zmeventnotification setup

Post by tlundste »

Well, after doing some more reading... what I've started to do has been abandoned, anyway, and it looks like he (dlandon) is working up a new, more consolidated install. I'll keep an eye on that one, and then try to figure out how to convert over to it. For now, I'll remove my "=1" entries, and go back to the much faster start up time.
Magic919
Posts: 1381
Joined: Wed Sep 18, 2013 6:56 am

Re: Stepping through zmeventnotification setup

Post by Magic919 »

He mentioned it in this thread viewtopic.php?f=41&t=30376 Abandoned might not be a good description.
-
tlundste
Posts: 19
Joined: Sun Nov 15, 2020 3:41 am

Re: Stepping through zmeventnotification setup

Post by tlundste »

Abandoned / Deprecated
To-ma-yto / To-mah-to.
No harm, or disrespect was intended in my comment... just that it's not a way to move forward anymore... no sense in struggling to configure something that isn't going to (likely) change with the times. Might as well wait for a more elegant solution (for now). I'm just playing, anyway... only got one camera.
tlundste
Posts: 19
Joined: Sun Nov 15, 2020 3:41 am

Re: Stepping through zmeventnotification setup

Post by tlundste »

I'm testing dlandon's .machine.learning docker image, on a VMWare workstation, Ubuntu installation running Docker. Eventually, I'd end up using it on a NUC7 running docker.

Both the new .machine.learning install running on my VMWare workstation, and the recently deprecated dlandon docker image running on my NUC7 seem to work pretty much the same, when giving respect to CPU resources... the VMWare instance is a bit slow, but it works.

Both installs will record my test camera, and alert on any motion in the frame (or motion in polygons in the frame if so configured). What I can't figure out what I need to do for this .machine.learning image (or the old base image running on my NUC7) is how to get it to alert on only people in the frame, or only dogs, or whatever. So.. can someone point me to the specifics of the documentation that bypass the "get it running" part, and hones in on the "get it alerting on known objects" part? Also, do I have to have any additional hardware (I've seen mention made of TPU processors on USB), or should it be able to work with just my virtual CPUs in my VMWare (or physical CPUs on my NUC7, supplied by Docker)?
Post Reply