STARCANADA 2017 Concurrent Session - New Test Techniques for New Technologies: Testing Computer Vision Applications | TechWell


Thursday, October 19, 2017 - 11:30am to 12:30pm

New Test Techniques for New Technologies: Testing Computer Vision Applications

Add to calendar

Computer vision capabilities are gaining ground in everyday commercial systems. Face recognition is now part of Windows 10; virtual reality games, gesture control, and autonomous cars are just some examples of the coming flood of computer vision products. However, many of the standard test techniques don’t work in these situations, so testers need to invent new ways to validate this technology. Join Michael Stahl as he shares how to test emerging computer vision capabilities. Michael provides an introduction to computer vision capabilities and shares demos of their applications. He then describes the difficult challenges facing testers of computer vision applications and outlines some approaches to address them, such as ground truth creation methods, selecting the right quality metrics, and modification of “classic” test techniques. Leave this session with some new ideas and be better equipped to test computer vision capabilities.

Michael Stahl
Intel Corporation

Michael Stahl is a software validation architect at Intel. In the last seventeen years Michael has tested code for cable modems, smart TVs, and graphics cards and is currently testing the code behind Intel’s RealSense 3D camera technology. In this role, Michael defines testing strategies and methodologies for test teams, delivers training and mentorship to junior engineers, and sometimes even gets to test things himself—which he enjoys most. You can view Michael’s papers and presentations at