Face and person detection with Deepstack - local and free!

I have installed both the object and face recognition.
When I try to run face recognition on the deepstack log I see /v1/vision/detection.
this is my configuration for the face recognition

  - platform: deepstack_face
    ip_address: 192.168.1.200
    port: 80
    api_key: !secret deepstack_api_key
    timeout: 5
    detect_only: False
    save_file_folder: /config/www/snapshots/
    save_timestamped_file: False
    save_faces: True
    save_faces_folder: /config/www/faces/
    show_boxes: True
    source:
      - entity_id: camera.camera_11
        name: face_counter

Can someone tell me what am I missing?

First, Iā€™m really thankful for the effort put in this.
My installation is only reporting false names. Iā€™ve tried to remove these two persons to start from scratch, but I canā€™t find a way to reset the object/ faces. Is there a way to delete everything Deepstack has learned? Maybe som file/ folder removal?

Great! Thanks for informing us. My question is why did it work in Lewis Video on YouTube? Is it just a matter of time until something has been fixed?

If you have deepstack installed in docker, then just install the container and do the cleanup. All data will be erased. Start the container and train it again.

It is difficult for me to answer your question. At the moment, I use the deepstack ui and Iā€™m happy with it.

@robmarkcole

Good day Rob, small question is there a way to prevent deepstack face from burning a timestamp on the processed image?
Iā€™m using in conjunction with this video and would like to remove it.

thank you

Hi guys,

I know it might sound like a silly questionā€¦ but how do I check if the image is being processed by Deepstack portainer? (I have home assistant running on ESXi and I deployed the deepstack using the portainer addon). at the begining I was trying to run in against my windows deepstack that I have for my blue iris setup but I donā€™t see any logs overthere (when I open the deepstack desktop shotrcut that suppose to show the processing).

Thanks,
Didi

For some reason, the timestamps are not written to my images

It keeps on using ā€˜Noneā€™ as a timestamp, hence the file is always overwritten.

This is how my configuration looks:

image_processing:
  - platform: deepstack_object
    ip_address: 192.168.116.2
    port: 5000
    save_file_folder: /shared/snapshots
    save_file_format: png
    save_timestamped_file: True
    always_save_latest_file: True
    source:
      - entity_id: camera.blink_voordeur

Anyone knows what might be wrong?

Found it myself - Iā€™m using Node-Red and while doing an object scan, I was also calling a camera.snapshot. That seemed to interfere with each other. Removing the camera.snapshot did the trick

watch out for syntax!!!

this will work:

service: image_processing.deepstack_teach_face
data: 
  name: "Michi"
  file_path: "/config/www/michi1.png"

Iā€™m trying to get this working but I am having issues (403)

Here is my portainer / docker logs

v1/restore
[GIN] 2021/06/17 - 08:54:47 | 403 |    8.700245ms |      172.17.0.1 | POST     /v1/vision/face/recognize
[GIN] 2021/06/17 - 08:56:20 | 403 |        68.6Āµs |      172.17.0.1 | POST     /v1/vision/face/recognize

This is HA


Logger: custom_components.deepstack_face.image_processing
Source: custom_components/deepstack_face/image_processing.py:228
Integration: deepstack_face (documentation)
First occurred: 9:54:47 (2 occurrences)
Last logged: 9:56:20
Depstack error : Error from Deepstack request, status code: 403

Iā€™ve used Portainer to manage the container

mapped 5000:5000
with 
VISION-FACE    TRUE
VISION-DETECTION    TRUE

And added this to my config

 - platform: deepstack_face
   ip_address: localhost
   port: 5000
   timeout: 5
   detect_only: False
   save_file_folder: /config/www
   save_timestamped_file: True
   save_faces: True
   save_faces_folder: /config/faces/
   show_boxes: True
   source:
    - entity_id: camera.camera_front_garden
      name: camera_front_garden

Iā€™ve also tried changing the host to the IP address - but no difference
Any ideas?

I was using this to seperate people from cars, etc. And give me different automation options. I have not been following for a while and looks like i have some reading to do. :slight_smile:

     person_in_driveway:
       friendly_name: Person in Drivway
       value_template: >
         {% set m = state_attr('image_processing.front_door', 'ROI person count') %}
         {{ m | float >= 1 }}

Obviously this does not work anymore. I realized some of my automations were failing. Anyone have a quick way to seperate counts based on target?

are sure that port 5000 is correct?
open a browser and point to the deepstack IP
for example: http://10.0.56.100:5000

do you get a website with deepstack image?

if not you are using the wrong port.
running it on nvidia jetson on 10.0.56.100 on port 80:

- platform: deepstack_face
    ip_address: 10.0.56.100
    port: 80
    timeout: 5
    detect_only: false
    save_file_folder: /config/www/faces
    save_timestamped_file: true
    save_faces: true
    save_faces_folder: /config/www/faces
    show_boxes: true
    source:
      - entity_id: camera.wohnzimmer_kamera
        name: face

yeah, no problem loading the webpage on 5000

I would really wanna know if you could train the colour detection / recognition. Have you found anything new on that?

Hi everyone,

I there a way to set file retention? to delete files older than x days?

Thanks!
Didi

You could use my custom component for this :https://community.home-assistant.io/t/imagelist-camera-integration-for-handeling-snapshot-files-camera-browse-move-delete-and-create-mp4-gif/272462

1 Like

I was actually looking at this earlier today :slight_smile: it looks great!! Iā€™ll give it a shot.
Thank you very much sir!

Didi

Hi guys,

more then 1200 commentsā€¦ What it is?!
Please answer me the simplest question: where is the managing UI? Just show me how to see the pictures of teached persons, how to change one of them if you have made and mistake in teaching?

PS
Hot ot at least remove a person?

Hi,

Iā€™m using Deepstack for a while now, but I keep getting low res pictures for my snapshots :confused:
Iā€™m using 4k cameras and getting horrible quality for the snapshots.
Iā€™m using BI and running the process_imaging on that (so no direct rtsp to HA) could that be a problem? (very few pictures are with full res)

Thanks,
Didi