The node(let) is based on the JetsonHacks' Jetson Nano CSI repository. As such, it only supports Raspberry Pi v2 cameras (and others that are supported by Jetson Nano). Additionally, it requires OpenCV with GStreamer support (see JetsonHacks' OpenCV build scripts to build one).
cap_width,cap_height(int, default: 1280, 720) - Width and height of the captured image. These parameters only affect the capturing mode! Be sure to set them to one of the supported modes for your camera (Raspberry Pi camera v2 supports 3280x2464@21fps, 3280x1848@28fps, 1920x1080@30fps, 1280x720@60fps).width,height(int, default: 640, 480) - Width and height of the output image.fps(int, default: 60) - Desired framerate/publish rate [Hz].flip_method(int, default 0) - Image rotation/flipping.0for no rotation, see detailed description at JetsonHacks.frame_id(string, default: "main_camera_optical")- Camera frame name.capture_delay(double, default: 0) - Additional time (in seconds) that is substracted from the timestamp. Allows to compensate for capture delays.topic_name(string, default: "image_raw") - Image topic name.camera_info_url(string, default: "") - URL pointing to the camera info yaml file.
The node is meant to run as a nodelet (jetson_camera/JetsonCameraNodelet). In fact, the node is basically a standalone simplified nodelet manager.
The image captured by this node may have a red tint that is more pronounced near the edges. This can be fixed to a degree by using an ISP camera override.
An ISP override file for an Arducam imx219 board (camera_overrides.isp) is placed in this repository in the extra folder. Place it in the /var/nvidia/nvcam/settings on your Jetson Nano, set its permissions to 664 and its owner:group to root:root.
Note that, according to RidgeRunner, only ODMs have access to the required tools for ISP calibration files.