Foundations of Computer Vision

Cloning (Large Repo)

⚠️ This repo is large!!

For a faster clone:

git clone --depth=1 git@github.com:Foundations-of-Computer-Vision/visionbook.git

Local Development

1. Install Quarto — download the CLI and verify with quarto --version

2. Preview the book:

   quarto preview

   This opens a live-reloading browser tab at localhost:<port>.

Content

Chapters are written in .qmd files (Quarto Markdown), which use syntax similar to Markdown with LaTeX math support. Quarto converts these to HTML for the website.

Figure Platform

Converts 2D textbook figures into interactive 3D HTML visualizations using GPT + Three.js, with automatic planning, generation, and critique scoring.

Quick Start

1. Install dependencies

# Backend
cd figure-platform/backend
npm install
cp .env.example .env        # then paste your OpenAI API key

# Frontend
cd ../frontend
npm install

2. Add your API key

Edit backend/.env:

OPENAI_API_KEY=sk-...your-key-here

3. Run

Open two terminals:

# Terminal 1 — backend (port 3001)
cd figure-platform/backend && node server.js

# Terminal 2 — frontend (port 3000)
cd figure-platform/frontend && npm start

Then open http://localhost:3000.

How It Works

1. Pick a chapter → see which figures are 3D candidates
2. Generate All → plans each figure, then generates interactive HTML (runs in parallel)
3. Auto-evaluate → critic scores each result on 5 rubrics (1–5) and flags failure modes
4. Results tab → browse by experiment, model, chapter; compare runs side-by-side