Beacon
Built on SD-Retinex, Beacon clarifies inspection footage from places light cannot reach, offline, on an existing PC without a GPU.
- Year
- 2026
- Role
- Lead, research and development
- Category
- App
- Status
- Active
- Stack
- SD-Retinex, Object detection, Local LLM
- Links
Beacon works with inspection footage shot where light cannot reach, such as sewer pipes and ship tanks. It started as a business plan that joins low-light image enhancement research with AI.
- SD-Retinex, training-free and CPU-only, corrects brightness and colour and brings out detail hidden in the shadows.
- Processing stays on the PC on site, with no internet connection, so footage never has to leave the organisation or go to the cloud.
- Detection and explanation with object detection and a local LLM are being built to run on site too.
The team is three people: Toma Okugawa, Uryu Den and Keito Inoshita.
Awards
- June 2026 — Tongali Business Plan Contest 2026: NICT Award, JR Central Award (Supporter Award), Beyond Next Ventures Award (Supporter Award) and Tongali Award, 5th place. The NICT Award came with a place at 起業家甲子園 (Kigyoka Koshien) in FY2026.