Image-Based Classification of Olive Varieties Native to Turkiye Using Multiple Deep Learning Architectures: Analysis of Performance, Complexity, and Generalization

TL;DR

This paper evaluates multiple deep learning models for classifying five Turkish olive varieties from images, analyzing their performance, complexity, and ability to generalize, with EfficientNetV2-S achieving the highest accuracy.

Contribution

It provides a comprehensive comparison of ten deep learning architectures for olive variety classification, highlighting the importance of parametric efficiency over model depth in limited data scenarios.

Findings

EfficientNetV2-S achieved 95.8% accuracy.

EfficientNetB0 offered the best accuracy-complexity trade-off.

Parametric efficiency is crucial under limited data conditions.

Abstract

This study compares multiple deep learning architectures for the automated, image-based classification of five locally cultivated black table olive varieties in Turkey: Gemlik, Ayvalik, Uslu, Erkence, and Celebi. Using a dataset of 2500 images, ten architectures - MobileNetV2, EfficientNetB0, EfficientNetV2-S, ResNet50, ResNet101, DenseNet121, InceptionV3, ConvNeXt-Tiny, ViT-B16, and Swin-T - were trained using transfer learning. Model performance was evaluated using accuracy, precision, recall, F1-score, Matthews Correlation Coefficient (MCC), Cohen's Kappa, ROC-AUC, number of parameters, FLOPs, inference time, and generalization gap. EfficientNetV2-S achieved the highest classification accuracy (95.8%), while EfficientNetB0 provided the best trade-off between accuracy and computational complexity. Overall, the results indicate that under limited data conditions, parametric efficiency plays a more critical role than model depth alone.