VGG Image Annotator
robots.ox.ac.uk
- Free plan
- Open source
- Windows
- macOS
- Linux
- Web
- Self-hosted
What VGG Image Annotator is
VGG Image Annotator (VIA) is a free suite of tools for manually annotating images, audio, and video. Its web applications are distributed as single HTML files that run offline in a browser without installation or setup. VIA3 supports images, video, audio, and shared or collaborative projects. VIA2 handles image files and remains maintained, but does not support shared projects. Users can mark regions with rectangles, circles, ellipses, points, polygons, and polylines, then import or export annotations as CSV or JSON. Scripts support conversion or export in COCO and Web Annotation formats, and a Python package converts VIA projects to COCO. Shared projects can use the Visual Geometry Group’s free public Version File Store, or users can run their own VFS. VIA is developed at the University of Oxford’s Visual Geometry Group and released under the two-clause BSD license for academic and commercial use. Download options include ZIP archives and standalone HTML copies for VIA2 and VIA3. The makers provide an in-app guide, FAQ, and email contacts for questions and feedback.
Who it suits
VIA suits researchers and other users who need to annotate images, audio, or video, including academic and commercial projects. VIA3 is the version to consider for shared projects.
Good to know, and what to watch
- Runs offline in a browser without installation.
- VIA3 supports images, video, audio, and shared projects.
- Exports annotations as CSV, JSON, and COCO.
- Released under a two-clause BSD license for academic and commercial use.
- Users can run their own Version File Store.
- VIA2 does not support shared projects.
- VIA2 supports image files only.
- No model-assisted labeling.
Plans
Every plan and priceFree
VGG Image Annotator
Free plan
The verdict
VIA offers offline annotation tools and multiple export formats at no cost, with VIA3 supporting collaborative projects. Users who need shared work should choose VIA3 rather than VIA2.
What the maker publishes
- Annotation attributes
- Region descriptions can use free text or predefined checkbox, dropdown, radio-button and image-list values. source
- Annotation features
- The user guide describes drawing regions such as rectangles, circles and points, and importing or exporting annotations as CSV or JSON. source
- Audio annotation
- VIA3 includes an audio annotator for marking speech or other time ranges. source
- Browser operation
- The applications run as standalone HTML files in modern browsers without installation or setup. source
- Browser requirements
- The project reports testing in Firefox, Chrome and Safari, with users also reporting success in other browsers. source
- Browser support
- The makers say they tested the application in Firefox, Chrome and Safari, and report that users have found it works in other browsers too. source
- Bulk editing
- The image-grid view supports bulk updating annotations across many images. source
- COCO interoperability
- Official scripts convert or export VIA annotations to COCO and other formats. source
- Dependencies
- The core software uses HTML, CSS and JavaScript without external libraries. source
- Documentation
- Documentation includes an in-app getting-started guide, FAQ, user guides, JSON-format documentation and per-version code documentation. source
- Downloads
- The page offers downloadable ZIP archives and standalone HTML copies for VIA versions 2 and 3. source
- Funding
- The project credits UKRI Visual AI and EPSRC Seebibyte research grants for development and maintenance. source
- Image annotation
- VIA supports manual annotation of still images for tasks such as object detection and segmentation. source
- Import and export
- VIA imports and exports annotation data in CSV and JSON formats. source
- Independent user praise
- A Reddit participant described VIA as their preferred tool because it was simple, self-hosted and fast. source
- Integrations and formats
- The page lists scripts for converting or exporting annotations in COCO and Web Annotation formats, and a Python package for converting VIA projects to COCO. source
- Intended users
- The page describes the software as suitable for academic projects and commercial applications, and includes tools for researchers annotating images, audio and video. source
- License
- VIA is released under the permissive two-clause BSD license for academic and commercial use. source
- License
- VIA is released under the permissive two-clause BSD license for academic and commercial use. source
- Lightweight size
- The full VIA application is distributed in a file smaller than 400 KB. source
- Limit: manual-only labeling
- The Roboflow tutorial says every box is drawn by hand and recommends AI-assisted tools for larger datasets. source
- Limit: older-version speed
- A Puget Systems review found VIA 1.0.x slow and required labels to be entered in a separate spreadsheet-like interface. source
- Limit: scale
- An independent guide describes VIA as suitable for quick tasks and small datasets rather than complex or large-scale projects. source
- Limit: unsaved work
- The Roboflow tutorial warns that VIA has no cloud storage or automatic saving, so closing the tab before exporting can lose work. source
Checked against the maker's pages on 21 Oct 2026.