Processing multiple RTSP streams and feedback via MQTT

[P.Kairies]

Hello!

I’ve been trying for over two weeks now to use a Raspberry PI 5 with Hailo8l to process multiple RTSP streams and then send a message to an MQTT broker. It should then receive something like “house/{camera name}/{person, face}” and create an image every x seconds to send to MQTT to see if the recognition works.

It just doesn’t work… I’m using Home Assistant on an N100 via ProxMox, and I’m really desperate. I’ve been thinking for a long time about whether I want to bother you with this, but unfortunately, neither ChatGPT, Gemini, nor Claude.ai can help me with my problem. They all provide completely bloated scripts that ultimately require new corrections, which then lead to another correction.

I’ve already tried Frigate, but unfortunately, it just doesn’t mount Hailo8 successfully. So it reverts to the CPU, and the Rasp5 starts whining!

Thank you in advance for any help you can give me!

Best regards from Germany

[omria]

Hey @P.Kairies,

Welcome to the Hailo Community!

I’d actually recommend going with Frigate - the Hailo integration is officially supported there. If you can share your config file and either your docker-compose file or the command you’re using, I’d be happy to help you get everything set up and running smoothly.

You can also check out this guide :

Hailo official integration with Frigate

Hey guys, I’ve been playing around with AI Hat+ on a pi5, and it seems to be working pretty nicely. However I’m at a loss as to what the different variants of the models are for. Is there a guide for this somewhere or could someone explain a bit further please?

[P.Kairies]

Thank you so much! It works now!