Automating garage functions based on the arrival of a person and a bicycle is a fantastic way to improve daily convenience. This build solves the problem of needing immediate, local object and face identification for smart home triggers without relying on costly cloud services. By installing Frigate on a UGREEN NAS, the maker provides a powerful, self-hosted solution for AI surveillance.
The maker deployed Frigate using Docker Compose on a system featuring an Intel Pentium Gold 8505 CPU and 16 GB of RAM. To ensure efficient video processing and AI analysis, the composition configuration was set up with device mappings enabling the Intel integrated GPU for hardware acceleration. The project is enhanced by using a YOLOv9 AI model, which offers more reliable detection of people, bicycles, and other objects than the standard model.
Integration with Home Assistant is seamless; the integration exposes entities like Bicycle count and Person count. The automation triggers the garage door opening when the maker and the e-bike are detected within the private driveway zone. The maker also set up specific detection parameters, such as a yellow ebike detection zone over the driveway and a red motion exclusion mask over the public street.
One area that remains open for future discussion is whether the model installation command functions directly within Home Assistant without requiring the use of a NAS and Docker setup. The build is a very impressive demonstration of local home automation capability.
Project by Smart-Live. Watch it on their YouTube channel: