Key points are not available for this paper at this time.
Lavender (Lavandula spp.) is a high-value essential oil-bearing plant cultivated extensively for its use in cosmetics, pharmaceuticals, and food products. Accurate identification of lavender species is essential for maintaining varietal integrity, ensuring uniform cultivation, and optimizing essential oil yield and composition. In this study, a deep learning-based approach has been developed to classify two commercially important lavender species—Lavandula angustifolia and L. latifolia—using a curated image dataset comprising 3,285 samples. Ten convolutional neural network (CNN) architectures were implemented and evaluated, including MobileNetV3, InceptionV3, Xception, VGG19, DenseNet121, ResNet101, and EfficientNet (B0–B3). The models were fine-tuned through transfer learning, learning rate adjustments, and batch size optimization. Additional layers were incorporated into baseline models to improve classification performance. Among all architectures, the modified VGG19 model achieved the highest accuracy of 100% on the test set, with reduced training time. The proposed framework demonstrates the applicability of deep learning in species-level classification, contributing to improved horticultural decision-making and supporting quality control in essential oil production systems.
Umar et al. (Wed,) studied this question.