Last Watch AI is a locally hosted application for creating if-then automations based on computer vision events. Last Watch can be used in tandem with IP camera software to trigger actions if the AI detects certain objects, like cars, people, or animals.
This project was heavily inspired by gentlepumpkin/bi-aidetection.
Last Watch watches for new image files in a configurable directory, then checks each image for a range of objects such as cars or people. Different automations can be set up to trigger if the AI detects relevant objects in the images.
A primary use case for Last Watch is to enhance Video Management Systems such as Blue Iris. For example, Last Watch can be set up to trigger security recordings only when a person is detected on security camera, thereby eliminating irrelevant motion events.
Last Watch can also be integrated with automation platforms suchs as Home Assistant, Node-RED, or openHAB. For instance, you can create virtual sensors which are set when a person is in a room, or when a car is parked in the garage.
- Platform Independence - everything runs in Docker containers
- Web Interface - desktop and mobile friendly
- Web API - everything can be managed with REST
- Offline and locally hosted
Supported Automations:
- Telegram - send images to bot
- Folder Copy - copy images to a local folder
- Smb/Cifs - upload images to a Samba share (Synology, Windows share, Home Assistant)
- Web Request - make http GET and POST requests
- MQTT - publish out MQTT messages
Last Watch has been tested on Windows, Debian, and Arch Linux. Dual-core CPU and 2GB of memory are recommended.
- Clone the repo
$ git clone https://github.com/akmolina28/last-watch-ai.git
$ cd last-watch-ai
- Create the .env file from one of the examples
$ cp configs/.env.linux .env
$ nano