Double Take

Double Take connects Frigate camera images to facial recognition services through a shared web interface and API. It's for people running camera systems and Home Assistant who want to identify familiar faces and use the results in home automations. The app runs on your own server in Docker and is open source under the MIT license.

Recognition uses separate backends: CompreFace, DeepStack, Facebox or Amazon Rekognition. Double Take can work with multiple detectors, so you can manage training images and review results in one interface across those services. The UI and API run locally; images go to the detectors you choose. Amazon Rekognition processes them in the cloud.

Its Frigate integration responds to camera events and analyzes snapshots, with repeated attempts to find a match that meets your confidence requirements. It can send recognized names back to Frigate as event labels. Match results and unknown faces can feed Home Assistant notifications and automations through MQTT, which also supports automatic sensor creation.

You can add or remove training images for named people, retain match images locally, and access saved images through the API with face boxes. Password protection covers both the interface and API. Other applications can submit images through the REST API, and scheduled pauses let you disable detection at set times. Double Take also supports OpenCV image preprocessing, Gotify notifications and Home Assistant add-ons. Docker images support x86-64 and ARM hardware.