Intel Corp. is stepping up its artificial intelligence play with the release of a new toolkit that will enable developers to deploy computer vision at the network edge for cameras and “internet of things” devices.

The chipmaker said that its Open Visual Inference & Neural Network Optimization toolkit is compatible with popular machine learning frameworks such as the open-source TensorFlow and Caffe.

OpenVINO is designed to work with Intel’s regular central processing units and also specialized accelerator hardware, such as its field-programmable gate arrays, which are computer chips that can be programmed for specific tasks. OpenVINO is also compatible with Intel’s Movidius vision processing unit, which was launched in February and is designed to speed up the execution of AI algorithms.

With OpenVINO, developers will be able to build and train AI models in the cloud and deploy them across a broad range of products. For example, a developer working for a retail firm could use the toolkit to deploy computer vision capabilities inside a range of edge applications at the point of sale, in security cameras or in digital signage.

“Processing high-quality video requires the ability to rapidly analyze vast streams of data near the edge and respond in real time, moving only relevant insights to the cloud asynchronously,” Tom Lantzsch, senior vice president and general manager of Intel’s IoT group, said in a blog post. “The OpenVINO toolkit is designed to fast-track development of high-performance computer vision and deep learning inference applications at the edge.”

Included in the toolkit are three new APIs: the Deep Learning Deployment toolkit, a common deep learning inference toolkit that scales across Intel Vision Products, and optimized functions for OpenCV and OpenVX.

Analyst Patrick Moorhead of Moor Insights & Strategy said OpenVINO was a welcome addition from Intel because it provides developers with “an easier IoT edge programming path across different accelerators,” including graphics, FPGAs and Movidius programmable ASICs. Moorhead had previously pointed out that FPGAs in particular are notoriously difficult to program, but said Intel was working to address this.

Intel said OpenVINO is already being used by a number of big enterprises, including GE Healthcare, which is using it for medical imaging applications. It can be downloaded now for free here.

Image: Intel