| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Original HTTPS Page] |
Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.
You must be logged in to block users.
Contact GitHub support about this user’s behavior. Learn more about reporting abuse.
Report abuseA static, five-page Astro + TypeScript website: Home, Publications, Experience, Awards, and CV. The blue-and-white design pairs a portrait-led homepage with compact experience cards and animated research capability previews.
Requires Node.js 22.12+ (Node 24 LTS recommended) and npm.
npm ci
npm run devOpen http://localhost:4321/AaronCIH/. The /AaronCIH/ base path is intentional: it matches the proposed GitHub Pages project site.
For a production preview:
npm run check
npm run build
npm test
npm run preview -- --host 127.0.0.1 --port 4321The tests inspect the built output, so run the build before npm test. Tests use Node's built-in runner; no test service is required.
The abstract concept illustrations have been replaced with actual project material. Full-size figure and source links are available in the research stories. This version does not include a live model-inference service; a demo URL can be added when a real hosted demo is available.
The carousel uses native horizontal scrolling and CSS scroll snap, not automatic slide rotation. Its arrows keep a pending target during smooth scrolling so repeated clicks advance one card per click instead of jumping an entire viewport. The left/right arrow keys work when the carousel region itself is focused, without intercepting native video controls. Reduced-motion preference makes card navigation instant.
Touch/trackpad scrolling, focus navigation, and resizing update the range and boundary buttons. Without JavaScript, all five cards and static media remain accessible in the horizontally scrollable region; nonfunctional buttons stay hidden. Thumbnail videos are not wrapped in links, so their native controls remain usable. Project titles and resource links provide navigation.
Each publication has a non-empty topics array, using one or more of:
The research-area selector chooses one category at a time. Papers appear under every assigned category, combined with the selected year; each paper is still listed only once. Home research cards, featured stories, and the publication index share the same tags. Formal paper titles are unchanged.
| Work | Categories |
|---|---|
| RAR | Image Generation and Editing; Multimodal Learning |
| UniRestore | Image Generation and Editing; Multimodal Learning; Visual Understanding |
| RobustVisRAG | Multimodal Learning; Visual Understanding |
| PDAF | Domain Generalization; Visual Understanding |
| RVSL / SJDL-Vehicle | Visual Understanding; Domain Generalization |
| APGCC | Visual Understanding |
| Semantic Guidance | Image Generation and Editing; Visual Understanding |
| Other restoration papers | Image Generation and Editing |
The 19 local assets comprise one portrait, 13 research images, and five organization logos. They were migrated at the user's request from the previous homepage, publications, and experience page.
The selected-research preview uses result demonstrations rather than architecture diagrams:
teaser-sources.json records provenance, durations, and poster timestamps. The original GIFs total over 42 MB; their four silent H.264 MP4 conversions total under 1.8 MB. FFmpeg converts them at 15 fps, at most 960 px wide, with CRF 24, yuv420p, no audio, and fast-start metadata. Full animation durations are preserved. Representative WebP posters provide a readable static alternative.
The converted files are committed in teaser assets; neither FFmpeg nor downloads from the original sites are needed to build or deploy. Videos have no initial src and use preload="none". They start muted when visible and pause offscreen, in background tabs, and while a full presentation dialog is open. A visitor's native-control pause is not overridden when visibility changes.
At the owner's request, research demonstration videos autoplay independently of the device's reduced-motion preference; this preference still suppresses decorative CSS transitions and smooth scrolling. Native controls provide an immediate way to pause motion. Browsers may block autoplay despite muting: in that case the player stays available with an explicit explanation and can be started manually. Actual media errors show a static poster and a separate error message. “View full figure” still opens the original research figure, while “Source” points to the teaser source.
| File | Purpose |
|---|---|
| research.ts | Publications, feature stories, topics, paper links, and YouTube IDs |
| profile.ts | Profile, experience, education, awards, and news |
| media.ts | Typed local image imports, alternatives, and source-page links |
| ProjectImage.astro | Optimized project figures and thumbnails |
| ExperienceTimeline.astro | Full Experience rows, consistently logo-left and text-right |
| ExperienceSummary.astro | Horizontal Home cards linking to full Experience entries |
| ResearchCarousel.astro | Five animated Home cards, single-card navigation, and responsive range controls |
| TeaserMedia.astro | Viewport-aware autoplay, native controls, and user-pause preservation |
| teasers.ts | Teaser posters, clips, captions, and source links |
| ResearchTopics.astro | Shared multi-category tags for cards, stories, and the publication index |
| ResourceLinks.astro | Optional resource buttons; absent links are omitted |
| global.css | Design tokens, responsive layouts, reduced-motion and print styles |
| Layout.astro | Common navigation, metadata, footer, and preview safeguards |
| astro.config.mjs | Site URL and GitHub Pages base path |
Use a YouTube video ID, not an embed URL, in video. Paper years and award dates should retain their actual precision. Optional links should be omitted rather than filled with #.
When replacing media, update both the local asset and its source record. The 32 tests cover routes, images, category membership, stable controls, news, experience links, teaser budgets, deployment configuration, and single-card carousel movement (including rapid clicks, native scrolling, resize, keyboard controls, and reduced motion). Playback tests execute the component's real client script with controlled media events to verify autoplay, automatic versus user pauses, autoplay denial, and media-error fallback. Browser checks additionally verify desktop scroll switching and autoplay with both reduced and normal motion settings, native pause/resume, and post-presentation scrolling.
Repository: https://github.com/AaronCIH/AaronCIH
Expected deployed URL: https://aaroncih.github.io/AaronCIH/
The deployment workflow installs with npm ci, checks types, builds, runs tests, and only then uploads/deploys the static artifact. Pull requests run validation without deploying. Node 24 is used in CI; no FFmpeg installation is needed.
Publishing requires authenticated write access to the repository:
gh auth login --web --git-protocol https --scopes workflow
git push -u origin mainIn the repository's Settings → Pages, select GitHub Actions as the source. Then run or rerun Build and deploy website in Actions:
gh workflow run deploy.yml --repo AaronCIH/AaronCIH --ref main
gh run list --repo AaronCIH/AaronCIH --workflow deploy.ymlIf the first push starts a deployment before Pages is enabled, rerun the workflow after changing that setting. Confirm the deploy job succeeds and test the root page plus /publications/, /experience/, /awards/, and /cv/ under the /AaronCIH/ prefix. Direct loads, images, clips, and refreshes must work; a successful Git push alone is not a verified deployment.
This is a profile preview, not a finalized academic CV. The workflow and URL above describe the deployment configuration; GitHub authentication and successful deployment must be verified separately.
Original profile source: previous personal website. Research descriptions were rewritten from the linked project abstracts. No paid MotionSites assets or templates are used.
A curated collection of papers & repos on Auto-Skill (self-evolving agents) and Auto-Rubric (rubric learning from preferences) for LLM alignment & customization.
CVPR26 - RobustVisRAG: Causality-Aware Vision-Based Retrieval-Augmented Generation under Visual Degradations
CVPR2026-Restore, Assess, Repeat: A Unified Framework for Iterative Image Restoration
JavaScript 1
CVPR25(Highlight)-Unified Perceptual and Task-Oriented Image Restoration Model Using Diffusion Prior
| Back | FazBrowse Home | New Git URL |