Tomato Ripeness Annotation Tool

This project is developed by Jiacheng Rong et al. from China Agricultural Univerisity.

This project provides a tool for annotating tomato ripeness using a simple graphical user interface (GUI) built with OpenCV. The tool allows the user to select regions of interest (ROIs) in images, compute hue-based ripeness levels, and record occlusion status.

Features

• Annotate images for tomato ripeness estimation by selecting regions of interest.
• Automatically compute ripeness levels based on hue values.
• Manual annotation of ripeness levels by pressing keys '1' to '6'.
• Compute overall ripeness percentage by averaging weights of selected regions.
• Record occlusion status of images with Yes/No options.
• Save annotations to text files for further analysis.

Installation

Requirements

• Python 3.x
• OpenCV
• Numpy

Install dependencies via pip:

pip install opencv-python-headless opencv-python opencv-contrib-python numpy

Usage

Command Line Arguments

• --image_dir: Directory containing images to annotate (default: data/test/images)
• --annotation_dir: Directory to save annotations (default: data/test/annotations)
• --max_side: Maximum size of the longest side when resizing images (default: 600)

Example:

python annotation_tool.py --image_dir=data/test/images --annotation_dir=data/test/annotations --max_side=600

Key Commands

• Left-click: Add a square region for annotation.
• Mouse wheel: Adjust the size of the square.
• r: Remove the last added square.
• 0: Compute hue and ripeness level automatically.
• 1-6: Manually assign ripeness levels.
• s: Compute ripeness percentage and proceed to occlusion input.
• d: Move to the next image.
• Esc or q: Quit the program.

Annotation Output

Annotations are saved as text files with the format:

<ripeness_percentage> <occlusion_status>

Where:

• <ripeness_percentage> is a floating-point value representing the average ripeness weight.
• <occlusion_status> is 1 for occluded and 0 for not occluded.

License

MIT License

Author

Jiacheng Rong