Annotation Correction Tool

A Tkinter GUI for fast, visual correction of frame-level behavior annotations in videos.

Features

• Load a video (mp4, avi, mov, mkv) and a CSV annotation file
• Visualize the video frame-by-frame with behavior annotations
• Check/uncheck behaviors using checkboxes or keyboard shortcuts
• Navigate using "Previous", "Next", or jump directly to a frame
• Apply selected behaviors to a range of frames
• Save corrected annotations to CSV

Prerequisites

• Python 3.7+
• Required packages:
  • opencv-python
  • pandas
  • numpy
  • Pillow
• Tkinter comes with standard Python installations

Install dependencies:

pip install opencv-python pandas numpy pillow

Usage

1. Save the script (e.g., annotation_correction_tool.py) in your project folder.

2. Run the tool:

   python annotation_correction_tool.py

3. In the GUI:

   • Fichier → Charger Vidéo: Select a video file.
   • Fichier → Charger Annotations CSV: Select your CSV (format described below).
   • Use navigation controls to browse frames.
   • Edit behaviors via checkboxes or shortcut keys.
   • Use "Appliquer aux frames sélectionnées" to apply changes to a range.
   • Fichier → Sauvegarder or Sauvegarder sous... to save your work.

Input CSV Format

The annotations CSV must have:

• Frame column (with frame numbers, usually starting from 1)
• One column per behavior; each cell is 0 (absent) or 1 (present)

Example:

Frame,Grooming,Feeding,Resting
1,0,1,0
2,1,0,1
3,0,1,0

Keyboard Shortcuts

• Each behavior is auto-assigned a shortcut (first available letter)
• Checkbox label shows the shortcut, e.g. Grooming [G]
• Press the shortcut key to quickly toggle that behavior

Screenshots

coming soon