Detects known objects by color in camera images and sustained mechanical sounds (e.g. a kitchen hood fan) from camera audio.
There is great object and face detection software out there, but sometimes AI detection is overkill or not suitable different types of objects. Swatch was created to create an easy to use API to detect the presence of objects of known color in expected places.
In this example you can see a cropped section of the street with a trash can. Then, using expected color bounds, the parts of the image that don't match the expected color are masked away. If a significant enough amount of pixels are left that match the color then it can be considered a true positive.
- REST API to be invoked by other applications
- Intuitive frontend UI to manage swatch and use the Color Playground to fine tune colors for the config.
- Audio monitors that listen to a camera's RTSP audio for sustained mechanical noise (e.g. a kitchen hood fan), distinguishing it from speech/music.
- Voice satellites that turn a camera's microphone (and, with UniFi Protect, its speaker) into a Home Assistant Assist satellite with a local wake word.
- HomeAssistant Integration
- HomeAssistant Addon
A config.yaml file must be created and mapped to /config
Check out GitHub Docs for example and more instructions on creating the config file.
Check out the Color Setup Docs for how to setup objects and fine the color values to use.
Check out GitHub Docs for the API reference.
Contributions are very much welcome!
For instructions on how to get started, see the contributing section
If you would like to make a donation to support development, please use GitHub Sponsors.