IGNF/hierarchy_labellisation

Project to test Rust/Wasm hierarchical image segmentation as tool to label aerial image for creating IA training dataset

Hierarchy Labellisation

This repository contains the code for an experimental hierarchical labellisation tool. It is developped in Rust and compiles to WebAssembly.

It was developped as an end-of-study project at EPITA. The project was supervised by Nicolas David at IGN.

Installation

Prerequisites

• Rust 1.67+
• wasm-pack 0.10.3+
• Node.js + Yarn or NPM (for the example)

Build

wasm-pack build

This will generate a pkg folder containing the compiled WebAssembly module. It can be imported in any Javascript project using a bundler such as Webpack or Vite.

The file /pkg/hierarchy_labellisation.d.ts contains the type definitions and exported functions. You can use them to interact with the module.

Usage

You can find a working example in the example folder. The example is a simple vanilla Typescript project bundled with Vite. It allows you to load a TIFF image and compute its hierarchical segmentation. You can then use the slider to change the segmentation level.

To run it, simply run the following commands:

cd example
yarn install # You can also use `npm install`
yarn dev # You can also use `npm run dev`

The example will be available at http://127.0.0.1:5173/.