EA-CNN: Enhanced attention-CNN with explainable AI for fruit and vegetable classification.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Fruit and vegetable misclassification leads to financial losses in retail.
• Existing CNN models for classification are complex, computationally expensive, and lack interpretability.
• Limited datasets and classes are often used, hindering real-world applicability.

Purpose of the Study:

• To propose an explainable AI (XAI) driven enhanced attention-CNN (EA-CNN) for accurate and efficient fruit and vegetable classification.
• To improve classification accuracy and reduce computational cost compared to existing models.
• To provide interpretable predictions for practical applications.

Main Methods:

• Developed an enhanced attention-CNN (EA-CNN) model incorporating a novel pooling technique and attention mechanism.
• Utilized the comprehensive Fruit-360 benchmark dataset (141 classes) for training and validation.
• Employed an XAI approach for interpretable prediction analysis.

Main Results:

• The EA-CNN model achieved 98.1% accuracy on the Fruit-360 dataset with fewer iterations than baseline models.
• Demonstrated superior accuracy and reduced computational cost compared to existing methods.
• Validated the model's generalization ability on the 'Fruit Recognition' dataset, achieving 96% accuracy.

Conclusions:

• The proposed EA-CNN offers an effective and reliable solution for fruit and vegetable classification in practical applications.
• EA-CNN provides accurate, efficient, and interpretable classification outcomes.
• The model exhibits strong generalization, robustness, scalability, and adaptability across different datasets.