Home Assistant integration for the tensorflow-lite-rest-server. Make sure the tflite server is running then configure this integration as below. This integration adds an image_processing entity whose state is the number of target objects identified, which defaults to person detection. This integration does not automatically scan an image, you must call the image_processing scan service to process an image

Place the custom_components folder in your configuration directory (or add its contents to an existing custom_components folder). Then configure object detection. Important: It is necessary to configure only a single camera per tflite_server entity. If you want to process multiple cameras, you will therefore need multiple tflite_server image_processing entities.

Add to your Home-Assistant config:

image_processing:
  - platform: tflite_server
    ip_address: localhost
    port: 5000
    # scan_interval: 30 # Optional, in seconds
    save_file_folder: /config/www/
    source:
      - entity_id: camera.local_file

Configuration variables:

• ip_address: the ip address of your tflite server.
• port: (Optional, default 5000 )the port of your tflite server.
• save_file_folder: (Optional) The folder to save processed images to. Note that folder path should be added to whitelist_external_dirs
• source: Must be a camera.
• target: (Optional, default person) The target object class.
• confidence: (Optional) The confidence (in %) above which detected targets are counted in the sensor state. Default value: 80
• name: (Optional) A custom name for the the entity.