BOSTON — Neurala Inc. today launched a new video annotation tool assisted by artificial intelligence for its Brain Builder platform. “Automated video annotation will significantly accelerate data preparation for neural networks, in turn helping organizations train and deploy AI faster,” the company said.

The tagging of images and video is essential to developing data sets for modeling and training AI applications. Neurala offers Brain Builder on a software as a service (SaaS) basis to help streamline the creation, analysis, and management of deep learning.

“Traditional approaches to AI data preparation are extremely time-consuming and labor-intensive, requiring large amounts of data that needs to be painstakingly and expensively annotated,” said Massimiliano Versace, co-founder and CEO of Neurala. “Our aim with Brain Builder is to lower that barrier to entry with easy-to-use annotation tools. With the addition of video annotation, we’re able to automate data preparation even further, helping organizations reduce time and cost of AI data preparation by at least 50%.”

Neurala’s patented and award-winning technology came out of research into neural networks for NASA, DARPA, and the Air Force Research Labs in 2006. In 2013, the company joined the Techstars program for commercialization.

“Everyone wants AI, but they don’t know why,” said Heather Ames Versace, co-founder and chief operating officer at Neurala. “The video annotation tool is part of the Lifelong AI technology stack and provides transparency.”

AI-enabled annotation saves time, increases productivity

Neurala’s new tool can learn iteratively as users tag people, objects, or defects in a video. After a user tags items of interest in the first frame, the tool automatically annotates the same items in subsequent frames, said Neurala.

For example, if five people entered a frame, they would all be automatically annotated after a user tagged the first frame with just one person. In comparison, the user would have to tag each person as he or she entered a frame, which would take significantly more time.

In addition, AI-assisted video annotation can increase speed the tagging process and increase productivity, Heather Ames Versace told Robotics Business Review.

For instance, users could annotate one frame of a 10-second video and get an output of 300 annotations, whereas with traditional annotation methods, users would need to manually tag 300 distinct images to get the same result, said Neurala.

“Explainability and trust start with data,” said Heather Ames Versace at the recent AI World conference. “By annotating and tagging data in less time, a team can do more rapid prototyping.”

Saving money with Brain Builder

“At the end of the day, it’s about helping organizations and developers build, train and deploy AI more efficiently and cost-effectively,” Massimiliano Versace said. “Neurala’s Brain Builder platform is already changing the game when it comes to how vision AI is built. And now, video annotation will expand the possibilities for accessibility and productivity even further.”

Brain Builder can also provide a substantial return on investment, Neurala said. With Brain Builder, organizations can annotate at $6,750 per hour of video, compared with $13,500 without it.

Neurala has posted a tutorial that outlines the process and benefits of tagging objects in videos with Brain Builder. It also explains how to train a semantic segmentation network using TensorFlow.

In addition, the tutorial walks viewers through the steps for training across multiple GPUs, which can further reduce the training time.