spatial-ai-depthai

On-device spatial AI inference runs by connecting an OAK-D camera and starting the depthai_ros_driver node to publish RGB, depth, and neural network outputs to ROS 2 topics for real-time perception.

Yes, you need a DepthAI device from the OAK-D family, along with ROS 2 Jazzy or Rolling and the depthai_ros_driver package to execute the stereo depth and spatial detection pipelines.

Spatial AI depth sensing uses stereo cameras to perform on-device neural inference and 3D localization, enabling autonomous navigation and obstacle avoidance without cloud processing, which reduces latency and CPU load.

Yes, DepthAI pipelines handle neural inference and stereo depth directly on the OAK-D device, offloading computation from the host CPU and reducing latency for ROS 2 robotic applications.

The depthai_ros_driver node publishes RGB, depth, and neural network spatial detection outputs to ROS 2 topics, providing the 3D coordinates needed for spatial mapping and perception pipelines.