New Custom Compontent - Image Processing - Object Detection - DOODS

Share your Projects!
613

How can i set to check what and if it detects something? No label? Label with confidence 0?

614

Use the label * with a confidence of 0

615

if i try to use the label * in config.yaml i receive this error:

name of an alias node must contain at least one character (343:15)
340 | file_out:
341 | - "/config/www/tmp/{{ camera_entity.spl …
342 | labels:
343 | - name: *
---------------------^
344 | confidence: 0

616

Try putting the * in double quotes

617

Hey, can anybody guide on how to customize the boxes? Yellow is almost not recognizable on my feed and the description is too small to read. I was trying to make the box customization code from GitHub work but only get an error from HA that boxes are not supported with DOODS.

Thank you :slight_smile:

618

I’ve tried DOODS for a little while, but decided to not to have any camera’s directly on my HAOS computer, since it’s just not quite powerful enough to run them.

I’ve de-installed the addon, but there’s some leftover files that I want to remove, but I’m clueless on where to find them… The file size is about 1.2GB, making my daily full HAOS backups unnecessary big.

Setup:
HAOS baremetal on X64 intel and installed DOODS as Addon

619

No idea if someone is reading this thread, but give it a try…

I’m running my homeassistant for a little while now, but not an expert.
I added an IP camera, and in homeassistant added with mjpeg camera and gives a nice view.
Now added Doods2 as addon and is running.
Added the image_processing part in my configuration.yaml as documented.

But testing with the service image_processing.scan I can not select a camera.
If I add the camera myself, an error in the log:
Referenced entities camera.logeerkamer are missing or not currently available
But the camera is there, a live view is in the dashboard and checked it’s entityid is coorectly in the configuration.yaml.

In teh doods webui I can paste the url of the camera and the detect stream button shows the right live view.

What am I doing wrong?

620

If you’re still looking… try checking here: hassio-addons/doods2 at master · snowzach/hassio-addons · GitHub I made a bunch of updates and the colors should work now assuming you’re using HASSIO.

621

I don’t really use HAOS… But I suspect it’s caching the DOODS docker image. You need to use something to delete old images. If you can run a docker command it’s usually docker system prune or if you have portainer installed, I think you can delete the images there.

622

I think it’s because your cameras don’t provide a still image feature. You can add cameras to Home Assistant but to run image processing it has to know how to get a still image from the camera.

If you look up the MJPEG camera (since it sounds like that’s what you’re using) MJPEG IP Camera - Home Assistant you can see there’s a still_url parameter. If you can add that, I believe it might fix your problem…

623

Thank you for the quick response!
It’s an old topic…

Well, both url’s are filled in from the beginning.
The still image url is working in my browser.
Also in the webui from doods it work when clicking the detect stillimage button.

624

@snowzach

i use dood2 as HA addon, but i dont get the entity with the name “image_” is this config still okay or must i change doods to doods2 ?

# doods AI - Personen erkennung 
image_processing:
  - platform: doods
    scan_interval: 4
    confidence: 60
    url: http://192.168.0.73:8080
    detector: default
    source:
      - entity_id: camera.esp32_cam
    file_out: /config/www/esp32_cam/aufnahme.jpg
    labels:
      - name: person
        confidence: 60

EDIT

nevermind it looks like its “doods” but i dont get the entity with the processing, called “image_xXXXX”

625

You have to

You have to create an image entity and read the file that you wrote out with DOODS

626

yeah but perhaps i misunderstand something. with this code i generate a image entity or not?

# doods AI - Personen erkennung 
image_processing:
  - platform: doods
    scan_interval: 4
    confidence: 60
    url: http://192.168.0.73:8080
    detector: default
    source:
      - entity_id: camera.esp32_cam
    file_out: /config/www/esp32_cam/aufnahme.jpg
    labels:
      - name: person
        confidence: 60
627

Can you please help ? Homeassistant doesn’t generate the entity with the name inage_processingxxxxx

I dint know what to do

628

image_processing only looks at the image… And the output is the number of things it sees… You are writing a file /config/www/esp32_cam/aufnahme.jpg so in order to get the image entity you need to create a static camera entity and point it at your file that you output…

camera:
    - platform: local_file
      name: aufnahme
      file_path: /config/www/esp32_cam/aufnahme.jpg

That will then create an entiry called camera.aufnahme that has the image written by doods.

1 Like
629

@snowzach is there any way to specify a minimum size for an object? So i can say detect person only when over a certain size?

630

Not currently… Sorry… :frowning:

633

Hi @snowzach, great job with Doods2, I’m running it now on Home Assistant Green using both default and tensorflow. As expected I find tensorflow much more accurate, but slower. I was reading up on pytorch but for the life of me I can’t seem to get it working. I’ve read it will auto-download the latest model(s), and that I can place models in my /share folder, but I can’t seem to get either to work? Am I missing an instruction somewhere?

NOTE: I’m using the default config after installing the add-on and get this in the log file:
doods.doods - ERROR - Could not create detector pytorch/pytorch: Unknown detector type pytorch.

Thanks for any help!

634

Unfortunately Pytorch doesn’t work on raspberry pi/arm64 etc… It requires too much in the way of resources…