Machine Learning-Based Tomato Fruit Shape Classification System.

Area of Science:

• Agricultural Science
• Computer Science
• Genetics

Background:

• Tomato fruit shape is crucial for quality, commercial value, and breeding programs.
• Current subjective visual inspection for classification is inefficient and error-prone.
• Standardized classification is needed for genetic studies and cultivar descriptions.

Purpose of the Study:

• To develop a robust, objective fruit shape classification system for tomatoes using machine learning.
• To establish a novel classification framework with improved accuracy and standardization.
• To overcome limitations of manual grading in breeding and research.

Main Methods:

• Trained and evaluated seven supervised machine learning algorithms on a public dataset from the Tomato Analyzer tool.
• Utilized existing classification systems as label variables for model training.
• Derived a new seven-class classification framework based on class-specific metrics.

Main Results:

• The Support Vector Machine (SVM) model demonstrated superior accuracy compared to human classifiers.
• The novel classification system achieved an average accuracy of 88%.
• The system maintained high performance on an independent validation dataset.

Conclusions:

• The developed machine learning system provides a standardized and accurate method for tomato fruit shape classification.
• This approach reduces bias associated with visual inspection, aiding genetic research and consumer preference studies.
• Implementation of this system will enhance consistency and consensus in tomato breeding and variety registration.