TAG — Taxonomy Annotation Gateway

A browser-based image crop labelling tool built for reviewing and annotating whole image classification datasets.

Features

• Import a flat or structured image folder; structured subfolders are detected as label suggestions
• CSV import — attach filename, label CSVs to create label groups; first row is always skipped as header
• Multiple label groups — each image can be labelled across independent groups (e.g. Material, Item Category, Brand...)
• Suggested labels — pre-imported labels shown as dashed pills; click to confirm
• Confirm suggestions on navigate — optionally auto-confirm suggestions when moving to the next image
• Filters — multi-select filter bar: All, Unlabelled, Flagged, or any label per group; filters combine with AND logic
• Group exclusions — per-label dropdown to exclude other groups from being required (e.g. "Some Litter items may not have a brand label")
• Stats panel — live per-group progress, label distribution, and export buttons
• Export — CSV or folder-structured zip, named TAG-Export-{group}-{timestamp}.zip
• Zoom overlay — click any image to open a full-screen pan/zoom viewer

Stack

• React 19 + Vite
• Material UI v9
• JSZip

Development

npm install
npm run dev