Deep learning-based analytics can help solve real-life problems for various vertical markets. One of them is the government sector, which can leverage the powers of AI in infrastructure inspection work.
Needless to say, artificial intelligence (AI) and deep learning have become a hot topic. In fact, deep learning-based analytics can help solve real-life problems for various vertical markets. One of them is the government sector, which can leverage the powers of AI in infrastructure inspection work.
That was a point raised by
Incubit, a Japanese video analytics developer which has taken its various solutions to Computex this year. “We’re a start-up based in Japan with a branch office in Taiwan. We focus on AI-, deep learning-based image recognition solutions. We help clients develop image recognition models with deep learning, using various data collected,” said Naoki Kitamura, CEO of Incubit.
One of the exhibits Incubit had on display at their booth in the Innovex section of Computex this year was a solution that detects cracks, peeling paint and broken paint on walls, roads, highways and other infrastructure types.
“Different companies have different requirements for inspection. For example, Japanese companies require inspection companies to detect cracks with 0.06 millimeters. The requirements are different for different entities or countries, and we can train the software based on those requirements,” Kitamura said.
How gov’t can benefit
According to Kitamura, one of the primary end users that employ this solution is government agencies that do inspection work.
“The end user is oftentimes the government, because they need to maintain infrastructure. The government may also outsource to infrastructure inspecting companies, and they may work with drone companies. We work with all of them,” he said. “Before, they were checking cracks manually, and they had to hire a lot of people to check if there was a crack. By introducing this technology, we can reduce human cost, increase the speed of inspection and increase its quality. That's the benefit.”
According to Kitamura, the software is trained with hundreds of images of cracks, peeling paint and broken paint. It is loaded on a server with an NVIDIA GPU that has enough processing power to ensure maximal effectiveness. Since the company is a pure software developer, it has teamed up with different hardware manufacturers to make the solution works.
“We have partnership with NVIDIA. We’re in their partnership program. We also work with drone companies,” Kitamura said. “A big market for drones is for inspection. So we can reduce the cost of collecting data by introducing drones. Before we needed to have people go up and down for this purpose. Now we can increase productivity by using drones.”
Also at their booth, Incubit demonstrated other deep learning-based algorithms for various vertical needs. One of them, for example, is a solution that supports agricultural robots, which can recognize tree branches that have grown too long. The robot will then cut them accordingly.
All the solutions reflect Incubit’s mission to help end users solve real problems with advanced technologies. “No matter how good the concept or plan is, unless the idea can be brought into the real world, it is useless. Our mission is to realize the passions of businesses and products in the best form to the real world,” Kitamura said. “We always stand by our clients as partners as we share their passions, troubles, and joys. With that in mind, at Incubit we transform imagination to inspiration using the best technology and design.”