Comparative Performance Analysis of MobileNetV2, ResNet50V2, and EfficientNetB0 Models Using Transfer Learning for Medicinal Plant Leaf Image Classification

Mohammad Yoga Firnanda; Rohman Dijaya; Ika Ratna Indra Astutik; Yulian Findawati

doi:10.32664/smatika.v16i03.2520

Authors

Mohammad Yoga Firnanda Universitas Muhammadiyah Sidoarjo
Rohman Dijaya Universitas Muhammadiyah Sidoarjo
Ika Ratna Indra Astutik Universitas Muhammadiyah Sidoarjo
Yulian Findawati Universitas Muhammadiyah Sidoarjo

DOI:

https://doi.org/10.32664/smatika.v16i03.2520

Keywords:

Convolutional Neural Network, EfficientNetB0; Image Classification, Medicinal Plant Leaf Classification, MobileNetV2, ResNet50V2, Transfer Learning

Abstract

Accurate identification of medicinal plant leaves supports the correct and safe use of herbal remedies, yet manual identification is often unreliable because many species share closely related leaf shapes, textures, and colors. Transfer learning has emerged as a practical strategy for image classification tasks of this kind, since it reuses the feature representations of pre-trained convolutional neural networks (CNNs) and thereby lowers both training time and computational demand. This study aims to compare the performance of three transfer learning backbones—MobileNetV2, ResNet50V2, and EfficientNetB0—on a five-class medicinal plant leaf recognition task covering guava (Psidium guajava), basil (Ocimum basilicum), aloe vera (Aloe vera), betel (Piper betle), and pandan (Pandanus amaryllifolius), using a working dataset of 1,750 leaf images (350 images per class before deduplication), and to statistically verify whether the performance differences among the three architectures are significant. Data were partitioned 80:10:10 for training, validation, and testing, and each architecture was trained across five independent runs under an identical configuration; Accuracy, Precision, Recall, and F1-score were used to quantify performance, followed by the Shapiro–Wilk and Kruskal–Wallis tests. MobileNetV2 produced the strongest mean Accuracy (98.27% ± 0.52), while ResNet50V2 led on Recall. The Kruskal–Wallis results, however, showed no statistically significant gap between the three architectures on any metric (p > 0.05). In other words, MobileNetV2, ResNet50V2, and EfficientNetB0 can be regarded as functionally equivalent for this classification task, and the practical choice among them should therefore weigh factors such as inference speed, model size, and deployment constraints rather than average predictive performance alone.

References

[1] B. Setiyono and others, “Identifikasi Tanaman Obat Indonesia Melalui Citra Daun Menggunakan Metode Convolutional Neural Network (CNN),” J. Teknol. Inf. dan Ilmu Komput., vol. 10, no. 2, pp. 385–392, 2023, doi: 10.25126/jtiik.20236809.

[2] R. Pujiati and N. Rochmawati, “Identifikasi Citra Daun Tanaman Herbal Menggunakan Metode Convolutional Neural Network (CNN),” J. Informatics Comput. Sci., vol. 3, 2022, doi: 10.26740/jinacs.v3n03.p351-357.

[3] S. Salsabila and A. Suharso, “Comparison of Deep Learning Architectures in Identifying Types of Medicinal Plant Leaf Images,” J. Appl. Informatics Comput., vol. 8, no. 1, pp. 39–46, 2024, doi: 10.30871/jaic.v8i1.6289.

[4] Y. Gulzar, “Fruit Image Classification Model Based on MobileNetV2 with Deep Transfer Learning Technique,” Sustainability, vol. 15, no. 3, 2023, doi: 10.3390/su15031906.

[5] M. A. Morid, A. Borjali, and G. Del Fiol, “A scoping review of transfer learning research on medical image analysis using ImageNet,” Comput. Biol. Med., vol. 128, no. 408, 2021, doi: 10.1016/j.compbiomed.2020.104115.

[6] H. J. Mustaqim and O. Suria, “Comparative Analysis of ResNet50V2, ResNet152V2, and MobileNetV2 Architectures in Monkeypox Classification,” J. Appl. Informatics Comput., vol. 9, no. 4, pp. 1326–1331, 2025, doi: 10.30871/jaic.v9i4.9756.

[7] S. Arnandito and T. B. Sasongko, “Comparison of EfficientNetB7 and MobileNetV2 in Herbal Plant Species Classification Using Convolutional Neural Networks,” vol. 8, no. 1, pp. 176–185, 2024, doi: 10.30871/jaic.v8i1.7927.

[8] D. R. Fauzi and G. A. Haqdu D, “Comparison of CNN Models Using EfficientNetB0, MobileNetV2, and ResNet50 for Traffic Density with Transfer Learning,” J. Intell. Syst. Technol. Informatics, vol. 1, no. 1, pp. 22–30, 2025, doi: 10.64878/jistics.v1i1.6.

[9] H. Kaur, R. Sharma, and J. Kaur, “Comparison of deep transfer learning models for classification of cervical cancer from pap smear images,” Sci. Rep., vol. 15, no. 1, pp. 1–18, 2025, doi: 10.1038/s41598-024-74531-0.

[10] D. J. Richter and K. Kim, “Assessing the performance of domain-specific models for plant leaf disease classification: a comprehensive benchmark of transfer-learning on open datasets,” Sci. Rep., vol. 15, no. 1, pp. 1–31, 2025, doi: 10.1038/s41598-025-03235-w.

[11] A. R, T. T, and A. Sudarshan, “Comparative Analysis of Deep Learning Models for Tomato Leaf Disease Classification Using Transfer Learning,” Int. J. Multidiscip. Res., vol. 8, no. 3, pp. 287–304, 2026, doi: 10.36948/ijfmr.2026.v08i03.80069.

[12] A. Sonita and others, “Herbal Plant Classification Using EfficientNetV2B0 Model and CRISP-DM Approach,” J. Tek. Inform., vol. 6, no. 4, pp. 2253–2262, 2025, doi: 10.52436/1.jutif.2025.6.4.5141.

[13] S. R. Yang, H. C. Yang, F. R. Shen, and J. Zhao, “Image Data Augmentation for Deep Learning: A Survey,” Ruan Jian Xue Bao/Journal Softw., vol. 36, no. 3, pp. 1390–1412, 2025, doi: 10.13328/j.cnki.jos.007263.

[14] W. Zheng, “Comparison of transfer-learning for lightweight pre-trained model on image classification,” Appl. Comput. Eng., vol. 53, no. 1, pp. 56–63, 2024, doi: 10.54254/2755-2721/53/20241244.

[15] M. Sandler, M. Zhu, A. Zhmoginov, and C. V Mar, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018. doi: 10.1109/CVPR.2018.00474.

[16] K. He, X. Zhang, S. Ren, and J. Sun, “Identity Mappings in Deep Residual Networks,” 2016, pp. 1–15. doi: 10.1007/978-3-319-46493-0_38.

[17] M. Tan and Q. V Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” 2019.

[18] R. D. C. N. Sahera and Y. Rahmawati, “Optimasi Penerapan Algoritma Convolution Neural Network dalam Klasifikasi Tingkat Kesegaran Daging Sapi,” SNTEI, pp. 207–212, 2022, doi: 10.37600/tekinkom.v7i1.1122.

[19] T. H. Pinem and Z. P. Putra, “Evaluasi Kinerja Algoritma Klasifikasi Deep Learning dalam Prediksi Diabetes,” J. Ilm. FIFO, vol. 17, no. 1, p. 17, 2025, doi: 10.22441/fifo.2025.v17i1.003.

[20] R. Sianturi, “Test Normality As a Condition of Hypothesis Testing,” J. Pembelajaran dan Mat. Sigma, vol. 11, no. 1, pp. 1–14, 2025, doi: 10.36987/jpms.v11i1.7091.

[21] A. Gunasti, R. Ariany, S. Duwi, and S. Ahmad, “Perbandingan Jumlah Kendaraan Roda 4 Pada Jalan yang Berbeda di Hari yangSama Menggunakan Analisi One Way Anova,” J. Smart Teknol., vol. 5, no. 3, pp. 422–429, 2024.

[22] Z. R. Mardhotillah, Bunga, and S. Rozi, “Tukey HSD Post Hoc Test untuk Perbandingan Karakteristik Lingkungan dan Sumber Daya Provinsi Provinsi di Indonesia,” J. Eng., vol. 3, no. 2, 2021, doi: 10.22437/jurnalengineering.v3i2.14445.

[23] I. E. Tampu, A. Eklund, and N. Haj-Hosseini, “Inflation of test accuracy due to data leakage in deep learning-based classification of OCT images,” Sci. Data, vol. 9, no. 1, pp. 1–8, 2022, doi: 10.1038/s41597-022-01618-6.