On the other hand, Geo-Tracer uses deep learning's Instance Segmentation and Adversarial Generative Network (GAN) techniques, depending on the features of the geographic object to be detected. Geo-Tracer uses both instance segmentation and adversarial generative networks. At its core, Geo-Tracer specializes in a wide area, and large volume land tracing and area calculation, which is difficult to do manually as outputs are required in a short period of time. In fact, the company also offers a Traffic Counter that measures the volume of automobile and pedestrian traffic.


In general, Orbital Net enables clients to use satellite imagery for generating maps and map elements using AI. “We are responding to the technological trend of satellite imagery and semantic segmentation with the technology of aerial photography and instance segmentation,” says Kobayashi. Here, the company can easily and accurately decipher small objects such as the shape of solar panels on a roof and determine the area of each building. Importantly, the company’s strategy is to use AI for content generation and as a tool to create high-speed, high-volume, and accurate data at once, which has been difficult to create with conventional human methods.

Kobayashi cites the success story of a client, an end-user, who was planning to do a site survey to identify buildings with solar panels, but the area was 6,000 square km making it practically impossible. Orbital Net studied the possibility of constructing the location information of the building where the rooftop solar panels were installed from aerial photographs using its "Geo Detector." Here, the company followed the principle that what can be read by a human can also be read by an AI, but if the training data is logically consistent with minimal wavering, it directly affects accuracy. As a result, the process of extracting 250,000 houses with solar panels on their roofs, georeferencing them, and installing the database was completed in six hours.

For long, Orbital Net has been performing aerial photo interpretation using AI. From now on, the company is focusing on providing a one-of-a-kind service by reading map elements, unique geographical features and storing them in the Location database. For example, in Japan, the national government, local governments, and companies are beginning to move toward the realization of a decarbonized society by 2050. To understand the current situation, Orbital Net has released a concentration map service based on the results of reading rooftop solar power generation facilities using AI to visualize the status of renewable energy installations throughout Japan. After expanding the database to a nationwide version, the company will officially release the map API to view the data on various map applications. In this way, Orbital Net will continue to build a database of AI-created geographies. “We would like to continue to develop services that create new value, rather than simply releasing such data to the public. Currently, we are positioned as a geospatial x AI solution vendor, but in the future, we plan to develop the market as an AI Created DATA provider,” concludes Kobayashi.