"CCTV That Once Only Recorded Footage Now Assesses Risks on Site"... DeepX Kicks Off AI Transformation of Road Infrastructure

Unveiled video analytics technology at the "ITS World Congress"
Enables license plate recognition and railway safety monitoring with ultra-low power consumption
Signed mass-production cooperation agreements with Baida and SotaTek

DeepX has unveiled ultra-low-power neural processing unit (NPU) technology that equips traffic-control closed-circuit television (CCTV) systems and roadside infrastructure with real-time artificial intelligence (AI) analytics.

DeepX said on October 22 that it had participated in the 2026 ITS World Congress, held in Gangneung, Gangwon Province, on October 19, where it presented traffic video analytics technology powered by ultra-low-power NPUs and domestic mobility deployment cases. During the five-day event, it also discussed ways to cooperate on infrastructure intelligence with overseas government officials and industry representatives.

An intelligent transportation solution based on an ultra-low-power neural processing unit (NPU) on display at the DeepX booth at the 2026 ITS World Congress. DeepX

The key to the technology presented by DeepX is its ability to process data on site. Conventional traffic cameras simply record video and transmit it to a central control center. But the situation changes when NPUs are embedded in cameras and roadside equipment. They can immediately detect vehicle and pedestrian movements and sudden accidents on site. They can also be connected directly to warning systems.

Applying AI to outdoor infrastructure such as ordinary roads and tunnels requires overcoming several constraints. Power consumption, heat generation, communications costs and data latency must all be considered, in addition to computing capacity. DeepX's chips perform high-speed computing using only 1 to 5 watts of power. By processing data at the source, they significantly reduce the amount of data sent to central servers, easing the burden on communications networks and reducing operating costs.

DeepX is also accelerating commercialization projects with major domestic mobility companies. It is pursuing the use of its technology in the Korea Expressway Corporation's smart tolling system to recognize license plates on vehicles using electronic toll collection (ETC) across multiple lanes. For a safety project with the Korea Railroad Corporation (KORAIL), the company is considering equipment in the form of edge boxes to inspect train pantographs and monitor workers on railway tracks. It is also discussing passenger-density analysis using cameras inside public transit buses with a domestic automaker.

Business partnerships with domestic and international companies have also followed. DeepX signed a memorandum of understanding (MOU) with Baida, a domestic radar specialist, to jointly develop object-recognition solutions for autonomous robots and vehicles. The companies will pursue field testing and mass production based on sensor-fusion technology. DeepX also signed an MOU with Vietnamese information technology (IT) company SotaTek to collaborate on solution development and mass production. It plans to target the Vietnamese market, focusing on traffic, security surveillance and robotics.

Kang Sanggyun, DeepX's managing director overseeing domestic sales, said, "As intelligent transportation systems become more advanced, technologies that analyze and make use of the vast amounts of data generated on site in real time will become increasingly important. Building on our experience applying technology in the transportation and mobility sectors with major domestic companies, we will expand the commercialization of ultra-low-power AI semiconductors in the global ITS market."

This content was produced with the assistance of AI translation services.