Machine Learning Turns Drones Into Literal Inspector Gadgets

We have written already this week about Amazon’s advances in automation, which could eventually replace the thousands of entry-level warehouse positions the company has created. When mobile app developers consider the issue of robots taking human jobs, they often think of this kind of warehouse or factory work. But as innovation in artificial intelligence and robotics accelerates, jobs in routine safety inspection and security are going automated. This makes sense: robots and drones are more easily able to access certain tough-to-reach areas (such as underwater pipelines), making them a cheaper and safer alternative for this kind of work. And now that deep learning has reached the point where sophisticated image recognition is possible, these robot inspectors have eagle eyes for details that the naked human eye could miss.

Boston-based Avitas Systems is one of the leading innovators in automated industrial inspection. The company deploys drones, autonomous underwater vehicles, and other robots to capture and analyze images of industrial structures like oil refineries and pipelines. Its fleet of robot inspectors is powered by Nvidia’s DGX-1 machine learning system, an off-the-shelf system available to any mobile app developer. DGX-1’s deep learning capabilities are advanced enough that the machines are able to recognize deviations in patterns of images. This means that the drones and other bots can detect even minor defects at these industrial sites.

According to estimates provided by Avitas Systems, utilizing these automated inspectors could save businesses one million dollars a year. Of course, this is good news for companies looking for cheap, reliable alternatives to actual human labor, but for the infrastructure inspectors and security guards out there, this is a pretty bleak development. Many Boston mobile app developers are excited about the utility and flexibility of drones, but automation remains a somewhat more controversial issue. It’s hard to not get pumped about what the combination of machine learning and robotics can accomplish in terms of efficiency, but many worry about the mass unemployment that could result from unchecked automation.