LabelMe Validator

A lightweight browser-based tool for quality-control and correction of LabelMe annotation files. Designed for reviewing cell-level segmentation masks, assigning corrected class labels per shape, and exporting results in both spreadsheet and LabelMe-compatible JSON formats.

• Browser: Chrome or Edge (recommended). Firefox works for basic use but does not support the 💾 Save Folder feature (File System Access API).
• Input: LabelMe JSON file(s) containing a shapes array. The imageData field (base64 image) is optional — if absent, only outlines are shown.

1. Open index.html in your browser.
2. Enter your validator name and press Start →. Your session is automatically restored on next visit.
3. Load annotation data:
   • 📂 Load JSON — drag & drop or click to open a single .json file.
   • 📁 Annotations Folder — select a folder containing multiple JSON files.
   • If served via HTTP, JSON files placed in the annotations/ subfolder are listed automatically.
4. Click a file from the list to open it.
5. Step through cells and assign labels using buttons or keyboard shortcuts.
6. Export your results when done.

Standard LabelMe JSON:

{
  "shapes": [
    {
      "label": "original_class",
      "points": [[x1, y1], [x2, y2], "..."],
      "shape_type": "polygon"
    }
  ],
  "imageData": "<base64-encoded image>",
  "imageWidth": 1024,
  "imageHeight": 768,
  "imagePath": "image.jpg"
}