API that allows for calls to an external computer running vision tasks (in our case, an NVIDIA Jetson AGX Xavier development kit) to speed up object detection inference and mapping tasks.
This is currently being used for the SUAS 2026 competition. Our ground station computer does not have a GPU, so we use the Jetson to offload the heavy computation.
- Clone this repository and set up the
uvvenv:
git clone git@github.com:MissouriMRR/inference_node.git
uv sync- Set the IP address you would like the API to listen on in
src/inference_node/main.py:
# This should be the IP address of the ground station on the interface the node is connected to.
HOST = "0.0.0.0"
PORT = 8642- Set up the systemd service:
sudo cp systemd/inference-node.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable inference-node
sudo systemctl start inference-node- Verify that the service is working by sending a request to the health endpoint on another computer:
curl http://{HOST}:{PORT}/health
{"status":"running"}