Labability - Image Label Tool

Motivation

I couldn't find a image label tool for my projects, so I decided to make one. This project is designed to label images, without any requirements for internet or self-deployed server.

Guide

Common Usage

1. Open the image folder you want to label
2. Press Alt while drag the mouse to draw a bounding box
3. Click on the box label to adjust the box
4. Double click anywhere outside the box to stop adjusting
5. Press Ctrl + Z to undo, Ctrl + Shift + Z to redo (not very stable)

Tips: On the right-top corner, you can resize the image to fit the window, or switch adjust mode.

Inferencing

1. Choose the YOLO model file you want to use in the tab with icon aimer
2. Configure the confidence threshold if needed
3. Press Detect button
4. Agree to detect (can be default agree in the checkbox below)

Configuration

1. Press the gear icon on the sidebar
2. Configure the object classes and their color/name

Roadmap

• label tool
• choose YOLO model file to inference
• use hardware acceleration for inference
• customizable object classes
• customizable classes color/name
• multiple export options
• optimize history(undo/redo) function
• optimize user interface

Build

yarn tauri build

Development

yarn tauri dev

Based on Tauri, Vite, Tailwind CSS