Training a machine to understand text is one thing, but training machines to analyze images can be something quite different.
AlchemyAPI has been working in the deep machine learning space, training neural networks to analyze information and carry out cognitive computing tasks.
While image recognition has had a steady place in its portfolio, the company on Tuesday launched AlchemyVision, an application programming interface specifically for visual data.
The system can take a photo, and using nothing but the pixel content, understand and map that image. AlchemyVision can recognize multiple objects and concepts within complex visual scenes, allowing computers to automate the task of tagging and classifying images into useful categories.
AlchemyAPI's CEO Elliot Turner said the real focus of the technology is enabling customers to take any visual content and assigning to a computer the process of sorting what is contained in that content.
"We have been doing research in computer vision for over a year and taken our history with brain based neural networks. For us as a company this ability was a natural evolution from our roots understanding text," Turner said. "We don’t want to distinguish between unstructured content — want our customers to derive value from all of it."
According to Turner, AlchemyVision represents the first API capable of acting upon visual information. He likened the technology to Google’s image search, and without knocking the Internet giant pointed out how AlchemyVision derives far more refined results.
Plus, Turner said, Google’s technology has remained locked internally, unlike AlchemyVision’s accessibility by any willing developer.