02/01/2026
Using IMX462 with Jetson Orin NX
Running a CSI-2 image sensor on an NVIDIA Jetson often sounds more complex than it really is. In practice, the setup is straightforward once the base system is prepared correctly. This guide shows how to bring up the Kurokesu IMX462 camera on the Jetson Orin NX using NVIDIA’s official tools and Kurokesu drivers. Full process is documented in video.
Preparing Jetson platform
Preparation starts with a bare Jetson Orin NX development kit. A jumper is installed to place the board into recovery mode, allowing it to be flashed from a host computer. Board is then powered using the barrel jack, while a USB-C cable connects it to a Linux host machine. On the host side, a standard Ubuntu 20.04 LTS system is used. NVIDIA SDK Manager is downloaded and installed, which handles the complete Jetson software stack. Through SDK Manager, JetPack 6.2.1 is selected and configured, then flashed directly onto the Jetson. This step prepares the operating system, CUDA stack, and core multimedia components in one pass. Once flashing is complete, power is removed and the recovery jumper is taken off. The Jetson is now ready for normal operation.
Installing camera driver
With the system prepared, Jetson is connected like a regular desktop computer: HDMI display, keyboard, mouse, Ethernet, the IMX462 camera, and finally power. The board boots into the freshly installed system. From the terminal, the IMX462 driver sources are cloned and compiled locally. This step integrates the camera into the Jetson media pipeline without requiring manual kernel patching. After installation, the camera is enabled using NVIDIA jetson-io configuration tool, where the IMX462 sensor is selected and basic parameters are applied. A reboot finalizes the configuration.
Verifying camera
After reboot, the camera is immediately available to the system. A simple GStreamer command is used to start a live preview, confirming that the sensor, driver, and CSI-2 interface are all working correctly. At this point, the camera behaves like a native Jetson sensor and can be used for streaming, recording, or computer vision workloads.