Advantech and Leopard Imaging are showing a machine-vision system that runs camera capture and AI inference locally on an edge computer. The demo combines four live camera feeds with person detection across eight additional video streams, without sending the video to the cloud.
Advantech Watch analysis
What happened
At AUSA 2026, the companies demonstrated Leopard Imaging cameras connected to Advantech’s MIO-5381 single-board computer, powered by Intel Core Ultra Series 3 processors. Each of the four live camera streams runs at 1920 × 1200 resolution and 30 frames per second. The demo also runs person detection on eight pre-recorded surround-view streams using OpenVINO and the computer’s integrated GPU.
Advantech says image processing and inference happen on the device, not in the cloud. The companies describe the partnership as an effort to make embedded camera integration simpler, with plans for broader camera support across computing platforms. Advantech’s announcement also describes image-quality testing and camera assessment for different lighting and operating conditions.
Why it matters
This is a concrete example of edge AI doing several jobs at once: handling multiple live camera feeds, decoding video and running detection on-device. That kind of setup could suit applications where sending video elsewhere is undesirable or where a compact system needs to do more than one thing at a time. The announcement does not establish performance in a deployed setting, but it does give developers a specific combination of hardware, cameras and software to examine.
Our read
The useful detail here is not simply that the partners have an AI demo. It is what the demo runs, on what hardware, and where the video stays. Local inference is a practical design choice, not a magic privacy guarantee; still, keeping processing on the device is a meaningful starting point for projects with real constraints around space, connectivity or data handling.
What to watch
- Whether the partners publish results from the demo under real-world operating conditions.
- Which additional camera and processor platforms they support.
- How image quality and system performance hold up across different environments.
Discussion spark: For camera-based AI systems, should local processing be a default requirement, or should teams decide case by case based on the application?
Sources and evidence
- Source update (11 October 2026, 16:00 UTC)
not affiliated with or endorsed by Advantech