How much does Rekognition cost if you’re off the free tier? I really don’t want to have to create new accounts every year just to keep the free tier if the costs aren’t horrible otherwise.
Honestly not sure. I’ve been using it for a few months and it hasn’t cost me anything. I’d only be looking at 30-50 submissions per day though.
Thanks! I failed to scroll further down the pricing page, looks like for the first million images analyzed it’s $1 per 1,000 images. So if you’re doing 50/day, it’s about $1.50 a month. I could live with that.
I’m doing something wrong somewhere - I followed all of the steps, but NodeRed is showing “API Error” under the snapshot and Rekoginition nodes. Is this Amazon’s API? HA’s API?
Does it give any details other than the API error? It’ll tell you what’s wrong with the call and highlight the node when you click on the error.
I’m getting this error:
"Call-service API error. Error Message: not a valid value for dictionary value @ data['entity_id']"
I’m guessing I took the instructions too literal. I used exactly “{{data.camera}}” for the Entity ID on ‘snapshot’.And I used {{data.image_processing}} for the ‘Rekognition’ node. Should I have specified cameras?
Here is a shot of my change node, along wi th the API errors on the side from ‘snapshot’ and ‘rekognition’. Not sure what I’m doing wrong. The change node looks correct to me.
:::EDIT::: Looks like I didn’t select ‘EXPRESION’ when setting the data. Just changed it to that. Maybe that will fix it. I’m not at home now, so I’ll have to wait for another movement on my cameras.
I can confirm that this fixed the API issue for me. However, I’m getting alerts from Pushover that show a car moving in my driveway, when I just want person detection. In my image_processing.yaml file, I have “target: Person” listed.
Yep, you’ll want JSONata for the type. It’s mentioned at the top of the node description. Could you post the contents of your image_processing.yaml file?
Here is my image processing yaml file. I didn’t merge them into separate YAML files per your tutorial, since that just simplifies things, right?
:edit: Maybe it was really late when I was setting this up and didn’t read it properly like I should have. I had the payload after the person check using the top exit node. Just re-read it and saw you use the bottom exit node. Would that be the issue?
Here is the image processing entity in HA:
That looks fine to me. Have a look at the section around the Person Check node. I just updated it along with the overall flow screenshot so it makes more sense. You should be using the top output based on the instructions above.
If it helps, I also labeled the section for each node so it doesn’t look so cluttered.
Thanks - yes it helps now. I changed it to the top exit node, and added the “>” than 0. However, you state to use number instead of string - which is fine - but it looks like you’re still using string in the example picture, fyi.
I just double-checked. That’s the numeric operator in the screenshot. Might need to refresh. I don’t recall if I’d uploaded it before I replied.
Thanks. I’m still using “string” as the state type, right?
A car just drove by my house - yet Amazon thinks it has a high percentage of a person being in the shot. I’ll probably change zones that it analyzes in Blue Iris, but is this normal?
Hrmmm. Let me play around with it. There’s probably a way to get a node to look at the confidence factor.
Or does it only send it to Amazon every so often? “Scan interval”? I haven’t had anyone walk by my house yet, but what if it detects say motion now. Then 2 minutes later? Will it not send the snapshot to Amazon the second time?
Either way, adjusting the confidence factor would be extremely helpful.
I think the interval is once a week just because it needs a value there. Realistically it should only trigger if there is motion. Also, I figured out how to get the confidence levels into the payload. Trying to figure out the best way to leverage them, although that might not be until later tonight. Work calls.
No worries. Thank you! 
Got it working :). I’ll try to post it tomorrow. You can set the confidence and it shows the percentage in the alert.