About Beacon

Beacon brings together Toma Okugawa, who researches low-light image enhancement at the National Institute of Technology, Yuge College; a researcher in large language models; and a behavioral-economics student who builds the business side.

It all started with research driven by one wish: that people and AI should be able to judge correctly even from footage captured in the dark. Founder Toma Okugawa published the fast Retinex methods FSSR and XCR in 2025 and SD-Retinex, which reframes their principle, in 2026. All three papers are public, and for SD-Retinex the code and evaluation results are open too.

Places where enough light cannot be brought in are everywhere in social infrastructure: sewer pipes, ship tanks, the inside of plant equipment. Building on the strengths of a technology that needs no training and runs on a CPU alone, we are creating software that draws reliable findings out of inspection footage from these sites.

As a team rooted in NIT Yuge College, a center of maritime education, we will take on field validation — starting with ships.