Face and person detection with Deepstack - local and free!

Share your Projects!
524

It is a WIP, but basically we will own it and can therefore get a tight integration with HA. I haven’t started a thread for this yet

525

random question as i imagine those doing this may be on this thread.
Anyone used an external GPU via thunderbolt on a NUC? wonder if that would be usable for this scenario with GPU

526

Finding a USB to mpcie Adapter seems not be an easy task. I mostly found USB to msata Adapters that probably won’t work.
Any suggestions for a suitable USB Adapter?

527

Robin, any chance deepstack_face would fire face_recognized event anytime soon?

528

I have a child on the way, and only limited time and too many projects. If there are feature requests I will proritise those where I am receiving sponsorship to cover those projects. My link is below, thanks:

529

Hi all
Before I bury myself in documentation and learning. Just checking, will my requirements be met?

  1. I have ring cameras. I can get video out (dont think images)
  2. I would like to detect and recognize faces. No requirement for this to be immediate
  3. Would like to keep videos that are not recognized, and delete videos that are

Reasonable?

530

I thought you couldnt get RTSP stream from Ring cameras. Is that possible now?

531

@ArnoNel face recognition is supported, but handling videos is not performed by this integration

532

Rob! Fantastic news… Congratulations with the little one :+1:

1 Like
533

Thanks Robin for all your excellent work! No more false alerts from security camera!!

Anyone know if there is a stand alone IP camera that has option to send command line like MQTT message when motion is detected and when motion event ends? I am currently using MotionEye with web camera but I would like to replace it with a IP camera that will also work as a motion sensor to trigger HA automation. I think ONVIF profile T supports that function but I don’t know which camera is compatible.

534

Checkout my article for links https://www.hackster.io/robin-cole/add-ai-brains-to-your-home-camera-system-5195b0

535

I am running deepstack docker container on a laptop with ubuntu OS. Can I use Intel Neural Compute Stick to speed detection so I can run multiple image processing with more cameras or does the stick only work with Pi?

536

I think its for the pi only. I would say try the deepstack forums

537

Anyone else seeing this after the latest Hassio update 0.103.4?

Platform error image_processing.deepstack_object - cannot import name ‘draw_box’ from ‘homeassistant.components.image_processing’ (/usr/src/homeassistant/homeassistant/components/image_processing/init.py)

538

They did some changes about image processing in the last update, you can see it in change log

539

I just released v2.5 which fixes the error introduced in HA 0.103. This will require users to be on HA >= 0.103

1 Like
540

As usual Rob… Always a pleasure!
Thanks for the quick reply.

Just updated and no errors :slight_smile:

1 Like
541

Thank You!
Works with HA docker container v 103.4 now.

1 Like
542

I’m trying the automation below but I get “Error while executing automation automation.new_person_alert. Invalid data for call_service at pos 1: extra keys not allowed @ data[‘file’]”

Any idea why? I basically copied the example from https://github.com/robmarkcole/HASS-Deepstack-object

- id: "1120092824611"
  action:
    - data_template:
        file: "{{ trigger.event.data.file }}"
        message: "Captured {{ trigger.event.data.file }}"
      service: notify.pushbullet
  alias: "New person alert"
  condition: []
  trigger:
    - platform: event
      event_type: image_processing.file_saved
543

I was able to fix it changing to

- id: "1120092824611"
  action:
    - data_template:
        data:
          file: "{{ trigger.event.data.file }}"
        message: "Captured Motion"
      service: notify.pushbullet
  alias: "New person alert"
  condition: []
  trigger:
    - platform: event
      event_type: image_processing.file_saved

Funny thing is that the image (“entity”_latest_person.jpg) is created but the event image_processing.file_saved does not trigger. In the developers tool I keep listening to that event and nothing shows up…

Ideas? I’m using 103.4 and the last deepstack version.