Engineering kits

Ready-to-use multi-camera systems based on NVIDIA Jetson

We present engineering kits for machine vision, that is a series of high-performance edge computers based on NVIDIA Jetson platform and equipped with machine vision cameras.

Various engineering kits with different Jetson modules are available off the shelf, including those built around Jetson Orin modules (AGX Orin, Orin NX and Orin Nano Super), as well as the latest Jetson modules from Thor family (Jetson T5000, T4000, etc.).

Each kit is fully operational, with a flashed Jetson module and the necessary software pre-installed. The software, which is based on MRTech IFF SDK, enables raw image acquisition from connected cameras, color pre-processing of the captured images on Jetson GPU and output of the resulting images in the required format to the destination using encoding and streaming, recording to a buffer or to file, rendering to the screen, among other operations.

If necessary, MRTech can customise both the hardware and software of the chosen system to address specific requirements.

The cost of the selected Engineering kit with Jetson Orin NX is determined by its hardware cost from authorized suppliers, plus MRTech service charge.

Engineering kit with Orin NX Super and PCIe cameras

How to connect XIMEA PCIe cameras to NVIDIA Jetson Orin NX?

Use XIMEA XEC-NX-3P-X2G3 carrier board for Jetson or

Choose MRTech Engineering kit based on Orin NX with connected XIMEA PCIe cameras.

Orin NX engineering kit with three XIMEA PCIe cameras connected.

We evaluated the performance of the engineering kit with Orin NX and three connected PCIe cameras (as shown in the photo) using a standard benchmark from MRTech, streaming images from all the cameras to the receiving computer.

Orin NX module ran the following image processing pipeline:

• Acquisition of raw images from three connected cameras.
• Color image pre-processing on Jetson GPU.
• H.265 encoding of the images using Jetson hardware accelerator.
• RTSP streaming.

The receiving computer had:

• Reception of three RTSP streams from a wired network or via Wi-Fi.
• Decoding.
• Rendering to screen.

Here are the test results:

• Image acquisition of 3x Full HD (1920×1080) at 120 FPS
• High quality color image processing on GPU at 1500 Megapixels/s speed
• H.265 encoder performance is 6x 1080p60 (747 MP/s)
• GPU + encoder processing latency of 13 ms
• Glass-to-glass latency (from camera sensor to receiving computer display) of 43 ms
• Power consumption of the Jetson module and cameras is 43W

Engineering kit with AGX Orin and PCIe cameras

How to connect XIMEA PCIe cameras to NVIDIA Jetson AGX Orin?

Use Connect Tech Forge carrier board and FireFly-M.2 adapters from XIMEA or

Choose MRTech Engineering kit based on AGX Orin with connected XIMEA PCIe cameras.

AGX Orin engineering kit with two XIMEA PCIe cameras connected.

We evaluated the performance of the engineering kit with Orin NX and three connected PCIe cameras (as shown in the photo) using a standard benchmark from MRTech, streaming images from all the cameras to the receiving computer.

Orin NX module ran the following image processing pipeline:

• Acquisition of raw images from three connected cameras.
• Color image pre-processing on Jetson GPU.
• H.265 encoding of the images using Jetson hardware accelerator.
• RTSP streaming.

The receiving computer had:

• Reception of three RTSP streams from a wired network or via Wi-Fi.
• Decoding.
• Rendering to screen.

Here are the test results:

• High-quality color image processing on GPU at 2000 Megapixels/s
• H.265 encoder performance is 2x 4Kp60 (995 MP/s)
• GPU + encoder processing latency of 25 ms
• Glass-to-glass latency (from camera sensor to receiving computer display) of 68 ms
• Power consumption of the Jetson module and cameras is 50W

Engineering kit with Jetson Thor and PCIe cameras

How to connect XIMEA PCIe cameras to NVIDIA Jetson Thor?

Use Auvidea X242 carrier board and PCIe host adapter or

Choose MRTech Engineering kit based on Jetson Thor with connected high-performance XIMEA PCIe cameras.

Thor engineering kit with two XIMEA PCIe cameras connected.

We evaluated the performance of the engineering kit with Orin NX and three connected PCIe cameras (as shown in the photo) using a standard benchmark from MRTech, streaming images from all the cameras to the receiving computer.

Orin NX module ran the following image processing pipeline:

• Acquisition of raw images from three connected cameras.
• Color image pre-processing on Jetson GPU.
• H.265 encoding of the images using Jetson hardware accelerator.
• RTSP streaming.

The receiving computer had:

• Reception of three RTSP streams from a wired network or via Wi-Fi.
• Decoding.
• Rendering to screen.

Here are the test results:

• Image acquisition of 2x 26.2 MP (5120×5120) at 60 FPS
• High quality color image processing on GPU at 3500 Megapixels/s
• H.265 encoder performance is 6x 4Kp60 (2986 MP/s)
• GPU + encoder processing latency of 18 ms
• Glass-to-glass latency (from camera sensor to receiving computer display) of 52 ms
• Power consumption of the Jetson module and cameras is 97W

Engineering kit with Orin Nano Super and MIPI cameras

How to achieve high performance from MIPI cameras connected to the Orin Nano module?

Use 4-lane MIPI CSI-2 interface with a suitable carrier board for Jetson and the right image processing software (MIPI driver and IFF SDK).

Order MRTech Engineering kit based on Orin Nano with MIPI cameras.

Orin Nano engineering kit with two IMX477 cameras with 4-lane interface.

We evaluated the performance of the Engineering kit with Orin Nano and two connected Arducam camera modules based on Sony IMX477 sensors (shown in the photo). As Orin Nano does not have a built-in encoder accelerator, we used for this a different standard benchmark with the following functionality:

• Acquisition of raw images from two connected cameras.
• Color image pre-processing on Jetson GPU.
• Image export to the AI application.

Here are the test results:

• Image acquisition of 2x 4K (3840×2160) at 60 FPS or 2x Full HD (1920×1080) at 240 FPS
• Color image processing on GPU at 1000 Megapixels/s speed
• Processing latency of 13 ms for 4K images and 5 ms for Full HD images
• Power consumption of the Jetson module and cameras is 28W