Industrial applications such as asset management and infrastructure inspection are among the roles being assumed by AI-based drones. Those aerial platforms are growing in sophistication with the addition of improved computer vision and deep learning technologies that can handle everything from dangerous infrastructure inspections to mundane asset tracking tasks.
Key AI chip players like Nvidia (NASDAQ: NVDA) have recently targeted the next generation of autonomous machines for industrial and manufacturing applications ranging from package delivery via drones to bridge inspections. Earlier this month, Nvidia released its Jetson AGX Xavier platform that includes development kit consisting of embedded AI processor and software stack.
Other vendors are jumping into the industrial market. For example, Boston-based Neurala said last week it is partnering with European utility asset management specialist Laserpas to provide its deep learning neural network technology. The partners said Neurala’s AI technology would be used to make drone inspections safer and cheaper while reducing the amount of time needed to analyze collected data through workflow automation.
Laserpas, Vilnius, Lithuania, said it would use Neurala’s Brain Builder AI workflow management tool designed to accelerate the creation of big data for deep neural network training. The platform’s annotation tools are intended to simplify and reduce the time need to tag static images and video collected by drones. Brain Builder is touted as providing “on-the-fly” learning to boost the accuracy of deep neural networks as more data is collected.
The tool also allows users to download data sets for deep neural networks using TensorFlow and the Caffe deep learning framework. Data also can be exported in different formats to train neural networks used for other applications.
Along with a analytics dashboard that runs across data sets and projects, Brain Builder runs either in the cloud or can be hosted on a server behind a firewall, Neurala said in announcing the drone inspection partnership.
Drone operators like Laserpas are increasingly relying on AI tools as they gather more image data from the field. “We are collecting pictures and videos containing critical infrastructure components that need to be carefully analyzed,” Mantas Vaskela, co-founder and CEO of Laserpas.
“AI-assisted data analysis is the only way forward for us.”
As chip makers like Nvidia target their AI chips at industrial robotics applications, Neurala and other machine learning specialists also have focused on improved techniques for training neural networks. In May, Neurala released a Neurala “lifelong” deep neural network software designed to boost the autonomy of robots, drones and other “smart” products. The framework utilizes a deep neural network pretrained on the ImageNet database of images organized by keyword. It also uses specific data sets required for specific tasks.
Neurala’s deep learning neural network software for drones and other smart devices is based on technology developed by the U.S. Air Force, NASA and the Defense Advanced Research Projects Agency.