Zoneminder | KI | Codeproject AI

[itoss]

Hi everyone,

I've been using ZoneMinder for over 10 years now and have generally been quite satisfied with it. However, after diving into AI-based video analytics, I started testing other NVR solutions—mainly Agent DVR and Shinobi—since they support AI image processing out of the box. I focused more on Agent DVR, as I really like the idea of having a dedicated AI server. This allows for a clean separation between the NVR software and the AI engine, which can improve compatibility and makes it easier to use AI server programs or APIs like CodeProject.AI. With this setup, multiple VMs can share a single passed-through GPU via the AI server, which is a big plus. Shinobi, on the other hand, integrates the AI components directly into the Shinobi VM, which feels more limited to me compared to Agent DVR.

That said, I recently discovered that both Agent DVR and Shinobi require a subscription for proper remote access or to use their full Web UI capabilities. In today's fast-paced software landscape, I'm really hesitant to pay for so-called "lifetime" support or features. Agent DVR is supposed to work with a reverse proxy if the Agent DVR cloud service is ever discontinued—but that’s not working properly in my case. When trying to access the Agent DVR web interface remotely (outside my LAN), I always get the message “no local ICE,” and I suspect that would be a permanent issue if the cloud service were shut down. A VPN solution like WireGuard could be a workaround, but a native, self-hosted web interface would be far more desirable.

Before I spend more time configuring Agent DVR, I’d like to give ZoneMinder another chance—especially regarding object, license plate, and face recognition. Unfortunately, most of the available tutorials are a few years old (2–4 years), and many are outdated or no longer functional. Ideally, I’d like to keep using ZoneMinder if it could record audio along with video (like Agent DVR does) and handle 4K streams.

Is anyone here running ZoneMinder 1.36 or 1.37 with object detection, facial recognition, and license plate recognition—ideally using the CodeProject.AI server? That AI server works great with Agent DVR, and as mentioned, it offers more flexibility with a passed-through GPU than running YOLO directly on the NVR VM.

Thanks in advance for any tips or suggestions!

[iconnor]

Despire their age, most tutorials should still be relevant.


In a month or two, 1.38 will get released and 1.39 will immediately get built in support for AI, including remote API based AI.