homebridge-face-location/README.md

# homebridge-face-location

Homebridge plugin providing real time location tracking via facial recognition.

## Installation

1. Download FFMPEG [here](https://ffmpeg.org/download.html)

   > Note: FFMPEG is required in order to run homebridge-face-location

1. Clone the repository by running `git clone ssh://git@thebword.ddns.net:3122/watsonb8/homebridge-face-location.git`
1. Run `npm install` to install required modules
1. Run `npm run build` to build the module
1. Run `npm link` to link this instance to your global homebridge instance

## Configuration

```
{
      "platform": "HomeLocation",
      "name": "HomeLocation",
      "refImageDirectory": "/path/to/reference/images/",
      "trainedModelDirectory": "/path/to/trainedModel/data.json/",
      "weightDirectory": "/path/to/trained/weights/",
      "outputDirectory": "/path/to/desired/output/image/directory",
      "trainOnStartup": false,
      "rate": 1,
      "detectionTimeout": 90000,
      "watchdogTimeout": 10000,
      "debug": false,
      "writeOutput": true,
      "rooms": [
        {
          "name": "Kitchen",
          "rtspConnectionStrings": [
            "rtsp://username:password@example.com"
          ]
        },
        {
          "name": "LivingRoom",
          "rtspConnectionStrings": [
            "rtsp://username:password@example.com"
          ]
        }
      ]
    }
```

#### Platform

- `refImageDirectory`: The location of the images used to train the facial recognition matcher

  > NOTE: This directory should only contain sub directories labeled with the name of the person to be matched with their corresponding images in the labeled folder. Any labels found in this directory will be used in face tracking

- `trainedModelDirectory`: The location of the trained `data.json` file if it exists. This is also the location where training data will be placed if `trainOnStartup` is true

- `weightDirectory`: The location of the pre-trained weight files used for face detection

- `outputDirectory`: The directory to be used if debug output images are wanted

- `trainOnStartup`: If true, a trained model will be created and persisted in the `trainedModelDirectory` during startup

- `rate`: The rate at which collect images from rtsp stream in fps

- `detectionTimeout`: The number of milliseconds to wait after a person is detected before returing to the undetected state

- `watchdogTimeout`: The number of milliseconds to wait before restarting the rtsp stream when a new image has not been recieved

- `debug`: If true, additional debug logging is displayed in the console

- `writeOutput`: If true, an image will be written to disk for every frame received from the rtsp stream

#### Rooms

- `name`: The name of the room in which to detect

- `rtspConnectionStrings`: A list of camera rtsp connection strings to associate with the room