C++ toolkit helps shorten the path to AI apps
- 27 March, 2017 21:00
Hoping to spread development of artificial intelligence applications, neural network platform builder Neurala is offering a toolkit that uses C++ for building AI applications.
The Neurala Developers Program includes C++ APIs, testing tools, code samples, and a working program with source code to integrate the Neurala Brain deep learning platform into applications. While mostly working with Linux, Neurala also offers wrappers for Google Android and Apple iOS application development.
Developers don't already need to know about AI in order to use the program, said Roger Matus, Neurala vice president of products and markets. They could, for example, build applications like the vision part of self-driving automobile applications; Neurala's software then enables smart products to learn and adapt in real time.
As an example, Matus cited an inspection application that can be deployed on drones, learning from its environment and classifying what it sees. "What our software can do is isolate the sections of the video where the inspectors should look," he said.
Neurala's commercially licensed platform features a software development kit and runtime; the kit works with processors from Nvidia, Intel, and ARM, and developers can access Neurala servers to run their applications.
The company's software differs from Google's Tensor machine learning library in that Tensor is an open source tool designed for researchers; Matus argued that it's not commercial-grade software. Neurala's software, on the other hand, was built for NASA and the U.S. Air Force and will be made available for the first time for commercial use.