cv-label

A desktop app for labeling image datasets for computer vision. Organize work into Projects → Tasks → Samples, and annotate images with bounding boxes or polygons in a canvas-based labeler with pan/zoom and per-label colors.

Built with Electron, React, and TypeScript. Data is stored locally in SQLite by default, behind an IDataStore interface designed to be swapped for other backends (e.g. an HTTP-based store) without touching any UI code.

Screenshots

Projects	Tasks	Samples

Bounding boxes	Segmentation	Annotations panel

Sample images are from the COCO dataset, used here for demonstration only.

Data formats

Import/export YOLO, COCO, and this app's own .cvlabel archive format — see formats/ for the .cvlabel spec.

Recommended IDE Setup

• VSCode + ESLint + Prettier

Project Setup

Install

$ pnpm install

Development

$ pnpm run dev

Build

# For windows
$ pnpm run build:win

# For macOS
$ pnpm run build:mac

# For Linux
$ pnpm run build:linux

Testing

Unit tests

Component and utility tests, run with Vitest + React Testing Library:

$ pnpm run test

End-to-end tests

Full app tests driven by Playwright, launching the real built Electron app (page objects live in e2e/pages/):

$ pnpm run test:e2e

Electron has no headless mode, so this opens real windows while it runs. On Linux CI (see .github/workflows/ci.yml), that's handled by running under Xvfb rather than by hiding the window, since Electron/Chromium throttle rendering for hidden windows, which makes tests slower and flakier, not faster.

CI

Pull requests and pushes to master run both test suites via GitHub Actions (.github/workflows/ci.yml). master is protected, so changes go through a PR and are squash-merged.