Hyperspectral Image Analysis
PhD research: analysis, recognition and restoration of hyperspectral images under challenging conditions, using deep learning to recover and understand scenes from hundreds of narrow bands.
Projects
Research directions, open-source tools, and things in progress.
PhD research: analysis, recognition and restoration of hyperspectral images under challenging conditions, using deep learning to recover and understand scenes from hundreds of narrow bands.
An open-source skill/plugin that gives a local AI coding agent (Claude Code, Codex, Cursor, OpenCode) a safe remote “arm” over SSH to read, run and diagnose a Linux server, while its brain, memory and credentials stay local. Built on native OpenSSH primitives with tiered safety.
A local-first prompt manager and template launcher for macOS: prompts are saved as “Arrows” with fillable slots, found by fuzzy search and recalled with one hotkey across AI chats, IDEs and terminals. Templates are plain Markdown, readable and shareable outside the app.
Master’s research: contact-free gesture recognition from WiFi channel state information (CSI). Proposed WiTransformer, improving generalisation, model size and robustness for real-world deployment.
Multi-angle, cross-scene image matching between drone and satellite views for UAV geo-localization; co-built the SUES-200 multi-height, multi-scene benchmark.
This website: built with Astro, in Chinese, English and Korean, combining an academic homepage and research notes.