the filename of the .jpg has changed in recent versions of the integration. The image you saw is likely the camera snapshot, not the .jpg with rekognition annotations.
The other thread with node red tutorial has the current jpg filename.
on mobile, dont have the url but easy thread to search for
@shizlkazizl you have listed labels as targets, but these are not objects. wood is a label, whilst wooden table would be an object. Check which objects are detected, then set those as targets.
1 Like
Anyone having amazon rekognition issues after upgrading to Home Assistant 0.113.0, i have reference of this in my configuration.yaml
image_processing:
- platform: amazon_rekognition
aws_access_key_id: !secret amazon_rekognition_access_key_id
aws_secret_access_key: !secret amazon_rekognition_secret_access_key
region_name: eu-west-1 # Optional region, default is us-east-1
save_file_folder: /config/www/amazon-rekognition/ # Optional, image storage location
save_timestamped_file: True # Set True to save timestamped images, default False
confidence: 80 # Optional, default is 80
targets: # Optional target objects, default person
- person
- dog
source:
- entity_id: camera.driveway
- entity_id: camera.driveway_sd_direct_ffmpeg
Log Details (ERROR)
Logger: homeassistant.config
Source: config.py:800
First occurred: 12:38:17 PM (1 occurrences)
Last logged: 12:38:17 PM
Platform error: image_processing - Requirements for amazon_rekognition not found: [‘pillow==7.1.1’].
Log Details (WARNING)
Logger: homeassistant.helpers.service
Source: helpers/service.py:188
First occurred: 12:40:50 PM (382 occurrences)
Last logged: 4:53:41 PM
- Unable to find referenced entities image_processing.rekognition_driveway
- Unable to find referenced entities image_processing.rekognition_driveway_sd_direct_ffmpeg
Works great with the doorbell. Maybe someday this can be part of the main add-on store. Until then, this works very well. Thanks
I think that drawing boxes only around object above confidence is not working. I have it set to 80 without any ROI in configuration and only person as targets and this is what I get:
and I get boxes in an image around person (that is correct) but also above object that is not person (confidene is only 43%)
My config:
- platform: amazon_rekognition
aws_access_key_id: !secret aws_access_key
aws_secret_access_key: !secret aws_secret_access_key
region_name: eu-west-1
save_file_folder: /config/www/image_processing
confidence: 80
source:
- entity_id: camera.kamera_home1
Hi guys, anyone having the same issue as me.
I run the service, the image is saved on the folder.
I run the service again and the image is not saving at all…
What do ya guys think?
Any plans to work on that with new release? 
Someone is/was working on an official integration for rekognition, so I have no reason to continue working on mine
If you’re talking about this https://github.com/home-assistant/core/pull/37784 , last comment isn’t very promising…
This is open source, anybody can contribute
Yeah. Wish I had skills needed…
Alright I did a major release adding several new features, check it out
1 Like
Thanks a lot! You are the Best.
1 Like
I don’t understand what the breaking change is. I updated and having issued with my target (person). Only change I saw was having confidence level - but that’s not required?
the config for targets have changed. If you are just using the default target (person) there is no change
Thanks - looked at the config again and had to add the “- target:”. Working again! Thanks!
First time user… playing around with this…no problem setting up and running . Thanks!!!
1 Like
Appreciate the support @wmaker As always feel free to propose improvements and new features on the repo