Automated Mango Variety Classification Using Deep Feature Extraction and Machine Learning Classifier Integration.

Area of Science:

• Agricultural technology
• Computer vision
• Artificial intelligence

Background:

• Manual mango classification is inefficient, leading to significant post-harvest losses in developing countries.
• Developing an automated system is crucial for improving efficiency and reducing economic impact.

Purpose of the Study:

• To develop a computationally efficient and highly accurate artificial intelligence framework for automated mango variety classification.
• To enable real-time applications in fruit sorting systems.

Main Methods:

• Evaluated eight deep transfer learning models as feature extractors.
• Combined these with ten classical machine learning classifiers.
• Assessed performance using accuracy, log loss, memory usage, training time, and inference latency.

Main Results:

• Hybrid models EfficientNetB0-Linear Discriminant Analysis (LDA) and ResNet50-Logistic Regression achieved 100% test accuracy.
• Inference time was reduced by up to 330 times compared to full Convolutional Neural Network (CNN) models.
• Demonstrated state-of-the-art accuracy with substantially lower computational cost.

Conclusions:

• Hybrid deep learning and machine learning architectures offer a viable solution for efficient and accurate automated mango classification.
• The developed framework is suitable for real-time applications and industrial fruit sorting.
• Future work includes real-world validation and embedded hardware deployment.