Qualcomm Technologies Inc. believes that mobile devices are the next frontier for deep learning, the artificial neural network technology behind recent advances in speech and image recognition. So today the company released its Neural Processing Engine software development kit on the Qualcomm Developer Network.

Qualcomm, which is the world’s largest mobile chip maker, said the NPE software development kit, or SDK, is the first deep learning software framework designed for its Snapdragon Mobile Platforms. According to the company, the new SDK will allow developers to provide on-device deep learning and artificial intelligence features on existing mobile hardware. The company suggested the SDK could be used for augmented reality, natural language understanding, gesture recognition and other applications.

“We developed the Snapdragon NPE SDK to help meet the increasing demand for running deep neural networks efficiently on mobile and other edge devices powered by Snapdragon,” said Gary Brotman, director of product management at Qualcomm. “The Snapdragon NPE was created to give developers the tools to easily migrate intelligence from the cloud to edge devices, and affords them the flexibility to target the appropriate Snapdragon core with the power and performance profile for a given user experience.”

According to Qualcomm, several companies have already adopted Snapdragon NPE to take advantage of deep learning on mobile devices. For example, Facebook Inc. plans to use the Snapdragon NPE to add augmented reality support to its mobile app. Qualcomm said in a statement that Snapdragon NPE gives Facebook five times the performance on the Adreno graphics processing unit versus a generic central processing unit implementation.

Smartphones are the most common devices to be powered by Snapdragon, but Qualcomm’s chips can be found in a wide range of products, including tablets, smart watches, digital media players, VR headsets and modems.

Qualcomm said Snapdragon NPE is designed to support custom deep learning programs, as well as common deep learning frameworks such as Caffe, Caffe2 and Tensorflow. The SDK itself includes runtime software, libraries, application programming interfaces, offline model conversion tools, sample code, documentation and debugging and benchmarking tools.

Photo: Qualcomm