Luxonis OAK-D-Lite Fixed Focus AI Stereo Depth Camera

The OAK-D-Lite is primarily programmed using Luxonis's open-source DepthAI API, which offers Python and C++ bindings. It is fully compatible with the DepthAI SDK for building custom perception pipelines. For roboticists, official ROS (Robot Operating System) 1 and ROS 2 drivers are available, allowing seamless integration of camera streams, depth data, and AI inference into robotic software stacks. Development typically occurs on a host computer running Windows, Linux, or macOS, which communicates with the device via USB.

The OAK-D-Lite is powered exclusively through its USB-C port. It complies with the USB 3.2 Gen 1 standard and requires a 5V supply. The device's power consumption can reach up to 5W under full load with all sensors and AI inference active. You can use any compliant USB-C port on a computer, a USB wall adapter, or a regulated 5V power source connected via a USB-C cable. Ensure the cable and power source can deliver sufficient current (typically 1A or more) for stable operation, especially when driving the high-speed data bus.

The Fixed Focus variant, as the name implies, has its RGB and stereo camera lenses locked at a specific focal distance. The RGB camera is fixed from 50cm to infinity, ideal for applications where the subject distance is relatively constant or for environments with vibration. The Autofocus variant features a moving lens element in the RGB camera, allowing it to dynamically focus from a very close range. Choose Fixed Focus for stability in mobile robots or drones; choose Autofocus for applications requiring sharp focus on objects at varying distances, like a handheld scanner.

Yes, the onboard RVC2 processor is designed to handle concurrent workloads. Its 4 TOPS of total processing power is partitioned, with 1.4 TOPS dedicated specifically to neural network inference. This allows the device to run object detection models (like YOLO or MobileNet-SSD) on the RGB stream while simultaneously calculating stereo depth from the two OV7251 cameras. All processing occurs on-device, outputting synchronized metadata (bounding boxes, labels), depth maps, and rectified video streams over the USB connection with minimal latency.