The rollout of AI-enabled chips, such as Intel’s Movidius line and platforms from GPU powerhouse NVIDIA, are proof positive that AI at the edge is more than simply a trend. It’s a seismic change affecting the machine vision sector.
The combination of very low-power, high-performance computing technology coupled with AI and deep learning is the key driver behind the newest embedded vision systems, according to Chris Yates, president of the European Machine Vision Association (EMVA).
These self-contained systems include camera sensors, along with image, vision,
and inferencing accelerators. Integrating all of these technologies into a single system without a common application programming interface (API) has been problematic for designers.
This prompted the EMVA, together with The Khronos Group, to announce the
formation of the Embedded Camera API Exploratory Group in February. Open to all
participants with an interest in embedded vision systems, the group’s first mission is to explore various use cases and requirements.
“We’re inviting people from many different industry backgrounds — companies
such as Intel, FLIR, and Stemmer Imaging — to participate at no cost,” Yates said.
“The first part is about alignment and open discussions about industry requirements.”
After the initial information-gathering phase, a statement of work will be drafted
for release later this year. No date has been set for the first release of an API, but
development is expected to start after the group has agreed on the scope.
The EMVA has a long history of shepherding standards, including GenICam and EMVA 1288, that are in widespread use today. Stay tuned for more developments.
In the meantime, I invite you to explore our current issue, highlighted by our
cover story “3D Vision Bolsters Robotic Bin Picking”. Find “Evaluating Illumination Options for Telecentric Imaging” by Edmund Optics’ Jeremy Govier. In “Neuromorphic Processing Set to Propel Growth in AI”,
Yole Développement’s Adrien Sanchez details the function and capabilities of the neuromorphic chips of tomorrow and their role in machine vision applications.
Enjoy the issue!
