Correct. My Windows desktop is on 24/7 and that’s where I run it. You could also run it headless on Windows and connect via the web interface.
1 Like
Hey! This looks like an awesome project
Thanks for sharing
2 questions before I set about it…
- Would this work in the Uk? (Not sure if any software that is maybe local to us only!?)
- Would this work with Ring cameras (doorbell and Floodlight)
Thanks again!!
For the people getting false positives… I noticed that Rekognition sometimes recognizes the ‘1’ in the timestamp that is rendered on top of the image as a person…
I’m working on something similar to this. The reason this does not work well for me is because I live next to a busy road and I want it to trigger on cars and persons. If it’s cloudy, I have a lot of shadow movements which will trigger rekognition. Probably it will detect a car on the road iso on my front yard. What I am implementing now is that it will mask out certain areas of the snapshot (e.g. the road).
I’m assuming it’d work in the UK. That shouldn’t make a difference. It should also work with Ring since it’s treated as a camera with a binary motion sensor.
Any idea how to push the picture/snapshot to a local file somewhere on HassIO? I know I already have the snapshots of any motion, locally, but I want to add the snapshot as a different name if Amazon detects a person. On my dashboard I want to have the latest snapshots of a person detected. I have it currently setup to show the latest motion, but that doesn’t help. Just can’t figure out how to output it from Node Red as well (I already have pushover - just want this in addition to it).
Great project. But I am having a weird issue which I hope someone can help me out with. It seems that the status of the Rekognition step in node-red does not return if a person is in image. It simply stays at sending until the person leaves the frame.
I suspected lag from AWS, but what is strange is that the rekognition entity in HA is being updated…
Has anyone seen this before?
I am using HA and node-red in docker under centos 7 if that makes any difference.
I haven’t, but it’s more likely a Home Assistant issue than Node-RED. The latter just makes an API call. What happens if you try to call the service directly from within HA?
What happens if you try to call the service directly from within HA?
Doing that from within HA actually worked fine.
As mentioned the image processing entity was getting updated correctly. It seemed almost as if Node-Red was making the API call but not getting the response back from HA and therefore would hang there.
Anyhow on a whim I upgraded HA to the latest release this evening, and now magically it seems to all be happily working together. Perhaps there was some sort of compatibility issue with HA 0.94 and the version 1.1 of the Rekognition component working with Node-Red.
I upgraded Node-Red to version 4.0.5 last night and now I am getting this error on the Rekognition node "Call-service API error. Error Message: not a valid value for dictionary value @ data['entity_id']". Anyone else seeing this, and how do I fix it?
Argh, just noticed that too. Let me take a look later and see what the issue is.
I’m not sure why, but just remove the value from “Entity Ids” in the Rekognition call-service node so it’s blank. It seems to grab the value from the payload automatically. I’ll look at it later this week and update the instructions.
Perfect! Works again, thanks.
Have a look at my edit to the Rekognition node. My comment above worked but was throwing warnings in the HA logs. This should be a bit cleaner. Still not sure what broke in the original one.
I’m having the same issue now. Seems like it’s working as long as the processing entity is empty. But if set, the rekognition call service node stays at state “sending” forever. I’ve updated HA to the latest version, but no change.
As mentioned by Morkin, it seems like the reponse from HA is not reaching Node-Red.
Any advise for this?
In my case the stuck sending turned out to be Node-Red crashing. This was not related to node-red, HA, or the Rekognition add-on, but was rather a malformed message going to Pushover causing a crash.
I did not immediately see this as the docker instance immediately restarted Node-Red. However in the logs it was plan to see.
Thank you! I noticed the crash in the docker logs. Problem was, that pushover can’t get the snapshot file from “/config/www/camera_snapshot/…”. Seems like it was in Node Red context because “/data” path worked.
I solved it with using the url of the snapshot file instead of the local path.
There was also an update to the HA module for Node-RED that referenced fixing something with call-service. I haven’t tried it but the related bug sounded similar. In the meantime, I’ve updated the instructions above for the Rekognition call that seems to correct the issue. I’ll try the old method at some point to see if it’s actually resolved.
My Node Red setup doesn’t trigger anymore…something with the latest update. I don’t get any notifications anymore…
Nevermind code error, fixed it…
How much hardware horsepower do you need for HA and this integration?