I recently came across an interesting discussion on using 10G Ethernet-based compact Holoscan Sensor Bridge (HSB) camera solutions with NVIDIA Jetson AGX Thor, AGX Orin, and IGX Orin platforms.
These setups are designed to work seamlessly with the NVIDIA Holoscan SDK, enabling real-time, low-latency transmission through a 10G Ethernet interface. What stood out to me is the integration of Lattice FPGA with a dedicated TintE ISP, which offloads image processing tasks from the GPU—allowing the Jetson to focus fully on AI inference and analytics.
Potential applications span medical imaging, robotics, industrial inspection, and research—any scenario where bandwidth, image quality, and compact form factor matter.
I’d be interested to hear from others experimenting with Holoscan-based cameras or 10G Ethernet vision setups on Jetson platforms.
-
Have you encountered latency trade-offs or bandwidth challenges?
-
How are you integrating ISPs or FPGA processing in your vision pipelines?
🎧 For a deeper dive, we recently discussed this in our Vision Vitals podcast:
