Projects
Research, apps and web work, newest first.
2026
- SD-Retinex
A training-free low-light enhancement method designed around the sigmoid derivative, fast on a CPU. The paper, code and evaluation results are public.
- Kyudo AI
A Flutter app that scores kyudo (Japanese archery) form with pose estimation that runs on the phone, and keeps hit records and club messages in one place.
- Beacon
Built on SD-Retinex, Beacon clarifies inspection footage from places light cannot reach, offline, on an existing PC without a GPU.
- Re-defining vanishing municipalities (IEEE GCCE 2026)
A framework that combines multidimensional clustering with small language models (SLMs) to re-define Japan's “vanishing municipalities” and derive policy insights.
- Why XCR and SD-Retinex work
A review site that checks, from ten perspectives and with derivations and reproducible experiments, why my low-light enhancement methods XCR and SD-Retinex work although they look as if they should not.
2025
- YUMI-TRACK
A web app that scores kyudo form from webcam video with pose estimation — the seed of Kyudo AI.
- XCR (eXponential-Cos Retinex)
A fast, structure-preserving Retinex method based on a complex exponential kernel, released as a preprint on Jxiv.
- FSSR (Fast Single-Scale Retinex)
A fast Single-Scale Retinex, evaluated for real-time human pose estimation (skeleton detection) under low-light conditions.
- Recipro Engine! (れしぷろえんじん！)
Cooking steps can be written as algorithms — a web app to learn home economics and programming at the same time.