Detect people in camera images using Sighthound Cloud. The Sighthound Developer tier (free for non-commercial use) allows 5000 images to be processed per month. If you need more processing per month you will need to sign up for a production account (i.e., a Basic or Pro account).

This integration adds an image processing entity where the state of the entity is the number of people detected in an image. For each person detected, an sighthound.person_detected event is fired. The event data includes the entity_id of the image processing entity firing the event, and the bounding box around the detected person.

If save_file_folder is configured, on each new detection of a person, an annotated image with the name sighthound_{camera_name}_latest.jpg is saved in the configured folder if it doesn’t already exist, and overwritten if it does exist. The saved image shows the bounding box around detected people and can be displayed on the Home Assistant front end using a Local File camera, and used in notifications. If save_timestamped_file is configured as true, then the annotated image is saved with a file name that includes the time of detection.

Note that by default the integration will not automatically scan images, but requires you to call the image_processing.scan service, e.g., using an automation triggered by motion.


To enable this platform in your installation, add the following to your configuration.yaml file:

# Example configuration.yaml entry
  - platform: sighthound
    api_key: "MY_API_KEY"
      - entity_id: camera.my_cam

Configuration Variables

api_key string Required

Your Sighthound Cloud API key.

account_type string (Optional)

If you have a paid account, used prod.

save_file_folder string (Optional)

The folder to save annotated images to.

save_timestamped_file boolean (Optional, default: false)

Save the processed image with the time of detection in the filename. Requires save_file_folder to be configured.

source map Required

The list of image sources.

entity_id string Required

A camera entity id to get a picture from.

name string (Optional)

This parameter allows you to override the name of your image_processing entity.

To verify the integration, check if a new entity is appeared as image_processing.sighthound_my_cam

Process an Image

When you want to process an image, you have to call image_processing.scan service and listen to the sighthound.person_detected and/or sighthound.vehicle_detected events.

An example using two automations:

  • The first automation is triggered, when a motion is detected. It calls the image_processing.scan service to send the camera image to the sighthound server for processing.

  • The second automation is triggered by a sighthound.vehicle_detected event. It sends a notification to a phone.

# Example automations.yaml entry
  alias: "Entrance Motion Image Processing"
  description: "Send a camera image to sighthound, when motion is detected at the entrance"
    - type: motion
      platform: device
      device_id: YOUR_DEVICE_ID
      entity_id: binary_sensor.my_motion_sensor
      domain: binary_sensor
    - service: image_processing.scan
        entity_id: image_processing.sighthound_my_cam
  mode: single

  alias: "Arriving Vehicle Notification"
  description: "Send a notification to a phone, when a vehicle is detected at the entrance"
    - platform: event
      event_type: sighthound.vehicle_detected
    - service: notify.mobile_app_my_iphone
        message: "Somebody has just arrived by car."
  mode: single