Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

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.


Features

Feature Details
Interactive cell viewer Zoomed-in canvas showing each annotated shape on the source image with a color-coded outline
5 built-in cell classes Assign labels via button click or keyboard shortcut (keys 1–5)
Multi-file workflow Load a folder of JSON files and switch between them without losing progress
Progress tracking Per-file progress bar and labeled/total counter
Cell list modal Browse, search, and jump to any cell by index, original label, or validator label
Session persistence Validator name and all progress survive page refresh
Auto-save Labels written to localStorage after every action; optionally mirrored to a .validation.json file
Export — Excel Full validation report (all files, all labels) as a .xlsx spreadsheet
Export — JSON Original LabelMe JSON with corrected label fields, ready for downstream use

Requirements

  • 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.

Quick Start

  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.

Keyboard Shortcuts

Key Action
1 steatosis
2 mesenchymal_cells
3 normal
4 non_nuclear
5 balloon_dystrophy
/ Previous / Next cell
Esc Close cell list modal

Input File Format

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"
}

About

Quality Control for LabelMe annotations

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages