I built a detection server that runs YOLO11 on Jetson Nano

[XuZhen86]

The 4GB memory on Jetson Nano is a bit too small for running Frigate, so I built this detection server that allowed me to move Frigate to a NAS (faster CPU and more RAM), while still using Jetson Nano (faster GPU) for detection. The server runs in a Docker container and Frigate can connect to it using the DeepStack config.

In addition, Frigate 0.16 is dropping support for Jetpack 4, which I assume the detection will stop working. Might as well getting it ready by decoupling the detection into a different project.

Please check out the repo and the details at https://github.com/XuZhen86/SimpleJetsonNanoDetectionServer.

Inference speed between 50 and 60ms when using the yolo11s model.