This is the web-based version of the Point Cloud Segmentation Annotation Tool (PSAT). It focuses on smooth visualization of large point clouds directly in the browser.

• Framework: React (with Vite for fast development)
• 3D Engine: Three.js
• Data Loading: loaders.gl (focused on .las and .laz support)
• State Management: Zustand
• Styling: Vanilla CSS

• Node.js: v18 or later (v20+ recommended)
• npm: v9 or later

1. Navigate to the web directory:

   cd web

2. Install dependencies:

   npm install

3. Start the development server:

   npm run dev

4. Open your browser at http://localhost:5173.

• Layout: The main layout is defined in src/App.jsx.
• UI Components: Simple UI components are used for controls and lists.
• Three.js Scene: Look into src/App.jsx for the Three.js initialization and point cloud loading logic.

This directory is a Git submodule. When working with it:

• Updating in main repo: If you make changes inside web/ and commit/push them to the PSAT-web repository, you must also go back to the root PSAT directory and commit the reference change.
• Cloning with submodules:

  git clone --recursive <main-repo-url>

• Pulling changes:

  git pull origin master

  And then update submodules:

  git submodule update --init --recursive

• High Performance: Renders hundreds of thousands of points smoothly using loaders.gl.
• Flexible Viewing: Toggle between RGB, Classification, and Instance views.
• Read-only Components: Display existing components from JSON/LAS metadata (no add/edit/delete in Web).
• Local privacy: All processing happens in your browser; no data is uploaded to a server.

• No polygon/box/lasso selection for per-point labeling.
• No per-point class/instance editing.
• No component add/edit/delete operations (view-only).
• No LAZ/LAS write-back from the browser.
• No category/instance visibility panels.
• Temporary: components with component_type = "绝缘子串" are hidden in Web display (3D markers and sidebar list).

• Point clouds: .las / .laz files.
• Metadata: Matching .json file (must be selected together with the point cloud due to browser security).
• ExtraBytes: Support for instance (uint32) and category (uint16) extra bytes in LAS/LAZ.

• JSON: Contains metadata and component points. Does not include point-level labels.