Answers are grounded in this paper and cite their sources.
The ripeness level of bananas significantly influences their shelf life, market value, and suitability for consumption. Traditional ripeness assessment methods are manual, subjective, and often inconsistent, leading to increased post-harvest losses and inefficient supply chain management. This study proposes an automated banana ripeness classification system using deep learning models, with a focus on Vision Transformers (ViTs), to analyze high-resolution images of Moira (Myndoli) bananas. The system categorizes bananas into five ripeness stages: Unripe, Midripe, Ripe, Overripe, and Rotten, based on visual features such as color, texture, and surface patterns.A dataset of 4,500 labeled images (900 per class) was utilized, with preprocessing steps including background removal, resizing to 224×224, and normalization. Controlled data augmentation techniques—such as horizontal flipping, rotation, and mild brightness and contrast adjustments—were applied to improve generalization while preserving color-sensitive features. Five deep learning architectures were evaluated: ViT-Base, Swin Transformer Tiny, CvT, ResNet50, and MobileNetV3. Model performance was assessed using accuracy, precision, recall, F1-score, and confusion matrix analysis under a stratified 5-fold cross-validation framework.Among all models, the Swin Transformer Tiny achieved the best performance, with an average cross-validation accuracy of 87.39% and a final test accuracy of 86.78%, demonstrating both high accuracy and stability. The selected model was further enhanced with Explainable AI (XAI) techniques, utilizing EigenCAM as the primary method and Grad-CAM for supplementary analysis, to generate visual explanations of predictions.The proposed system provides a reliable and interpretable solution for automated fruit grading, offering practical benefits for farmers, vendors, and supply chain stakeholders.
No takes yet. Share an insight, caveat, or question.
Arabekar et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: