Object detection with Amazon Rekognition

Hmm I suggest checking the HA repo for any similar sounding issues

Sorry, I wasn’t clear earlier. My questions was if you knew of any memory leaks in this component. Thanks.

I don’t think so, this behaves as any other image processing component, but please let me know if you conclude otherwise

when i was using this component, the memory usage kept going up into multiple GBs. when I stopped using it the memory usage came back normal (~250mb). I am not a developer so dont want to draw uninformed conclusions based out of my limited testing. If and when you get memory leak concerns from other seasoned users, you may look into this. cheers.

edit: never mind. it is still going up. may be something else causing this.

Is there a way to set a minimum confidence level for the sensor to turn on?

Great job on this. I’d previously written this guide for use with TensorFlow. However, I’ve been playing around with Hass.io which doesn’t easily support it. Recreated the flows using this plug-in without any issues. Great job!

1 Like

@cpuram thanks for flagging this, I will do some testing myself

@roofuskit you would need to use a condition in an automation

never mind. it is still going up. may be something else causing this.

just noticed on 92.1, this component is not loading and no error/warning messages in the log either. am I the only one?

All please update to release v1.1 when upgrading to HA 0.92 and above.
Thanks

2 Likes

thanks mate. updated and seems to work again now.

1 Like

An example config for 4 cameras and automation example would be greatly appreciated to get this beginner up and running. I would like to detect vehicles and people when motion is detected on any of the cameras.
Thank you in anticipation…

HI,
Wanted to ask if this would be possible to use this somehow to detect a U.S. Postal vehicle (kinda like a jeep but all white with red and blue stripes across the middle), namely as it stops-at/drives-by my mailbox in front of my house?

That is very similar to a problem I want to solve - in my case detecting a specific bus. I think this will require a custom model e.g. https://github.com/ahmetozlu/vehicle_counting_tensorflow

Hi Rob,

this works perfectly well. Setup took me 4 minutes including AWS key generation and HA restart :slight_smile:

Any chance of saving locally the detection picture, maybe with the boxes like TensorFlow?

Thanks a lot for the great work!

It sounds simple but requires changes to the way image processing works in HA. There’s a discussion about it in the HA architecture repo

Its 5000 per month, so if I did the math right its free around 1 scan per each 9 minutes, correct? (1 month is max 31 days * 24 hours * 60 minutes=44.640 minutes / 9 is 4960

1 Like

My cameras detect motion but my image_processing stays as “Unknown”. I think I am not getting set Keys correctly, tried several times but nothing works.

There is this error:
WARNING (MainThread) [hassio.updater] Can’t fetch versions from https://s3.amazonaws.com/hassio-version/stable.json: Cannot connect to host s3.amazonaws.com:443 ssl:True [SSLCertVerificationError: (1, ‘[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: certificate is not yet valid (_ssl.c:1056)’)]

OK now I’m getting this error:

You are using a custom integration for amazon_rekognition which has not been tested by Home Assistant. This component might cause stability problems, be sure to disable it if you do experience issues with Home Assistant.

But the states never change from unknown…

@ThaNerd you need to get to the bottom of that SSL error. The state unknown just indicates that the component has not processed any images.

1 Like