Research Article
Open Access
Comparative Study of Lightweight Deep Learning Models for Greenhouse Tomato Leaf Disease Classification Using the Public TLID Dataset
1
Wenzhou Academy of Agricultural Sciences, Wenzhou 325006, China
Corresponding Author:
Wei Luo,
[email protected]
Volume 2, Issue 1
2026
Received: 13 February 2026
Accepted: 17 March 2026
Published: 21 March 2026
Article Information
Abstract
Accurate recognition of greenhouse tomato leaf diseases is crucial for crop monitoring, timely intervention, and yield protection. In greenhouse environments, disease symptoms are often affected by complex illumination, background clutter, overlapping leaves, mixed patterns, and subtle inter-class differences, making reliable image-based diagnosis challenging. To evaluate compact convolutional neural networks for this task, this study presents a controlled comparison of five CNN models—MobileNetV3-Large, ShuffleNetV2\_x1\_0, MobileNetV2, EfficientNet-B0, and ResNet18—using the public Tomato Leaf Image Dataset (TLID). A curated split of 15,254 images covering seven conditions was used, with 10,674 for training, 2,286 for validation, and 2,294 for testing. All models were trained from scratch under identical preprocessing, augmentation, optimization, and selection protocols. Performance was assessed using accuracy, macro-precision/recall/F1, class-wise metrics, confusion matrix, and Grad-CAM visualization. Results show ResNet18 achieved the best overall performance with 71.80% test accuracy and macro-F1 of 0.6726. Among lightweight models, EfficientNet-B0 delivered the strongest results, reaching 70.40% accuracy and macro-F1 of 0.6299, establishing it as the most competitive lightweight baseline. Class-wise analysis indicated Healthy was recognized most reliably, while WhiteFly remained the most challenging due to limited samples, subtle cues, and overlap with healthy leaves. Grad-CAM visualization confirmed the best model focused on symptom-relevant regions rather than background. Overall, findings provide a benchmark for TLID-based classification, identifying EfficientNet-B0 as the strongest lightweight baseline and ResNet18 as the top-performing reference model.
Graphical Abstract
Keywords
greenhouse tomato leaf disease classification
lightweight convolutional neural network
TLID
EfficientNet-B0
ResNet18
Grad-CAM
Data Availability Statement
The TLID dataset used in this study is publicly available at https://doi.org/10.17632/kt64b2kh89. The final processed data split, trained model weights, and source code are available at the public GitHub repository: https://github.com/nilihao2003/tomato-leaf-classification.git
Funding
This work was supported in part by the Wenzhou Basic Scientific Research Project under Grant GG20250197, and in part by the Zhejiang Provincial Science and Technology Commissioner Project under Grant 2025CNYJY04.
Conflicts of Interest
The authors declare no conflicts of interest.
AI Use Statement
The authors declare that no generative AI was used in the preparation of this manuscript.
Ethical Approval and Consent to Participate
Not applicable.
References
- Mohanty, S. P., Hughes, D. P., & Salathe, M. (2016). Using deep learning for image-based plant disease detection. Frontiers in Plant Science, 7, 1419.
[CrossRef] [Google Scholar] - Hughes, D., & Salathé, M. (2015). An open access repository of images on plant health to enable the development of mobile disease diagnostics. arXiv preprint arXiv:1511.08060.
[Google Scholar] - Zimmermann, G. B., Pellenz, M. E., Costa, Y. M. G., & Britto Jr, A. S. (2025). Enhancing disease and pest detection in greenhouse tomato cultivation using advanced machine learning on new dataset of images. Journal of the Brazilian Computer Society, 31(1), 187-202.
[CrossRef] [Google Scholar] - Pellenz, M., Zimmermann, G. B., Britto Jr, A. S., & Costa, Y. M. G. (2025). Tomato Leaf Image Dataset (TLID/PTLID) [Dataset]. Mendeley Data, V2.
[CrossRef] [Google Scholar] - Brahimi, M., Boukhalfa, K., & Moussaoui, A. (2017). Deep learning for tomato diseases: Classification and symptoms visualization. Applied Artificial Intelligence, 31(4), 299-315.
[CrossRef] [Google Scholar] - Tan, L., Lu, J., & Jiang, H. (2021). Tomato leaf diseases classification based on leaf images: A comparison between classical machine learning and deep learning methods. AgriEngineering, 3(3), 542-558.
[CrossRef] [Google Scholar] - Zhang, Y., Wu, G., & Shen, J. (2026). Precise tea leaf disease detection using UAV low-altitude remote sensing and optimized YOLO11 model. PLoS One, 21(2), e0342545.
[CrossRef] [Google Scholar] - Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. C. (2018). MobileNetV2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 4510-4520).
[CrossRef] [Google Scholar] - Howard, A., Sandler, M., Chen, B., Wang, W., Chen, L. C., Tan, M., ... & Le, Q. (2019, October). Searching for MobileNetV3. In 2019 IEEE/CVF International Conference on Computer Vision (ICCV) (pp. 1314-1324). IEEE.
[CrossRef] [Google Scholar] - Ma, N., Zhang, X., Zheng, H. T., & Sun, J. (2018, September). ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design. In European Conference on Computer Vision (pp. 122-138).
[CrossRef] [Google Scholar] - Tan, M., & Le, Q. (2019, May). Efficientnet: Rethinking model scaling for convolutional neural networks. In International conference on machine learning (pp. 6105-6114). PMLR.
[Google Scholar] - He, K., Zhang, X., Ren, S., & Sun, J. (2016, June). Deep Residual Learning for Image Recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 770-778). IEEE.
[CrossRef] [Google Scholar] - Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE International Conference on Computer Vision (pp. 618-626).
[CrossRef] [Google Scholar] - Attallah, O. (2023). Tomato Leaf Disease Classification via Compact Convolutional Neural Networks with Transfer Learning and Feature Selection. Horticulturae, 9(2), 149.
[CrossRef] [Google Scholar] - Bhujel, A., Kim, N. E., Arulmozhi, E., Basak, J. K., & Kim, H. T. (2022). A lightweight attention-based convolutional neural networks for tomato leaf disease classification. Agriculture, 12(2), 228.
[CrossRef] [Google Scholar] - Chen, H., Wang, Y., Jiang, P., Zhang, R., & Peng, J. (2023). LBFNet: A Tomato Leaf Disease Identification Model Based on Three-Channel Attention Mechanism and Quantitative Pruning. Applied Sciences, 13(9), 5589.
[CrossRef] [Google Scholar] - Saeed, A., Abdel-Aziz, A. A., Mossad, A., Abdelhamid, M. A., Alkhaled, A. Y., & Mayhoub, M. (2023). Smart Detection of Tomato Leaf Diseases Using Transfer Learning-Based Convolutional Neural Networks. Agriculture, 13(1), 139.
[CrossRef] [Google Scholar]
Cite This Article
APA Style
Ni, L., Ye, F., Cui, X., Peng, X., Song, S., & Luo, W. (2026). Comparative Study of Lightweight Deep Learning Models for Greenhouse Tomato Leaf Disease Classification Using the Public TLID Dataset. Digital Intelligence in Agriculture, 2(1), 45–53. https://doi.org/10.62762/DIA.2026.103152
Export Citation
RIS Format
Compatible with EndNote, Zotero, Mendeley, and other reference managers
TY - JOUR AU - Ni, Lihao AU - Ye, Fuyin AU - Cui, Xiaojun AU - Peng, Xiaoman AU - Song, Shaoshuai AU - Luo, Wei PY - 2026 DA - 2026/03/21 TI - Comparative Study of Lightweight Deep Learning Models for Greenhouse Tomato Leaf Disease Classification Using the Public TLID Dataset JO - Digital Intelligence in Agriculture T2 - Digital Intelligence in Agriculture JF - Digital Intelligence in Agriculture VL - 2 IS - 1 SP - 45 EP - 53 DO - 10.62762/DIA.2026.103152 UR - https://www.icck.org/article/abs/DIA.2026.103152 KW - greenhouse tomato leaf disease classification KW - lightweight convolutional neural network KW - TLID KW - EfficientNet-B0 KW - ResNet18 KW - Grad-CAM AB - Accurate recognition of greenhouse tomato leaf diseases is crucial for crop monitoring, timely intervention, and yield protection. In greenhouse environments, disease symptoms are often affected by complex illumination, background clutter, overlapping leaves, mixed patterns, and subtle inter-class differences, making reliable image-based diagnosis challenging. To evaluate compact convolutional neural networks for this task, this study presents a controlled comparison of five CNN models—MobileNetV3-Large, ShuffleNetV2\_x1\_0, MobileNetV2, EfficientNet-B0, and ResNet18—using the public Tomato Leaf Image Dataset (TLID). A curated split of 15,254 images covering seven conditions was used, with 10,674 for training, 2,286 for validation, and 2,294 for testing. All models were trained from scratch under identical preprocessing, augmentation, optimization, and selection protocols. Performance was assessed using accuracy, macro-precision/recall/F1, class-wise metrics, confusion matrix, and Grad-CAM visualization. Results show ResNet18 achieved the best overall performance with 71.80% test accuracy and macro-F1 of 0.6726. Among lightweight models, EfficientNet-B0 delivered the strongest results, reaching 70.40% accuracy and macro-F1 of 0.6299, establishing it as the most competitive lightweight baseline. Class-wise analysis indicated Healthy was recognized most reliably, while WhiteFly remained the most challenging due to limited samples, subtle cues, and overlap with healthy leaves. Grad-CAM visualization confirmed the best model focused on symptom-relevant regions rather than background. Overall, findings provide a benchmark for TLID-based classification, identifying EfficientNet-B0 as the strongest lightweight baseline and ResNet18 as the top-performing reference model. SN - 3069-3187 PB - Institute of Central Computation and Knowledge LA - English ER -
BibTeX Format
Compatible with LaTeX, BibTeX, and other reference managers
@article{Ni2026Comparativ,
author = {Lihao Ni and Fuyin Ye and Xiaojun Cui and Xiaoman Peng and Shaoshuai Song and Wei Luo},
title = {Comparative Study of Lightweight Deep Learning Models for Greenhouse Tomato Leaf Disease Classification Using the Public TLID Dataset},
journal = {Digital Intelligence in Agriculture},
year = {2026},
volume = {2},
number = {1},
pages = {45-53},
doi = {10.62762/DIA.2026.103152},
url = {https://www.icck.org/article/abs/DIA.2026.103152},
abstract = {Accurate recognition of greenhouse tomato leaf diseases is crucial for crop monitoring, timely intervention, and yield protection. In greenhouse environments, disease symptoms are often affected by complex illumination, background clutter, overlapping leaves, mixed patterns, and subtle inter-class differences, making reliable image-based diagnosis challenging. To evaluate compact convolutional neural networks for this task, this study presents a controlled comparison of five CNN models—MobileNetV3-Large, ShuffleNetV2\\_x1\\_0, MobileNetV2, EfficientNet-B0, and ResNet18—using the public Tomato Leaf Image Dataset (TLID). A curated split of 15,254 images covering seven conditions was used, with 10,674 for training, 2,286 for validation, and 2,294 for testing. All models were trained from scratch under identical preprocessing, augmentation, optimization, and selection protocols. Performance was assessed using accuracy, macro-precision/recall/F1, class-wise metrics, confusion matrix, and Grad-CAM visualization. Results show ResNet18 achieved the best overall performance with 71.80\% test accuracy and macro-F1 of 0.6726. Among lightweight models, EfficientNet-B0 delivered the strongest results, reaching 70.40\% accuracy and macro-F1 of 0.6299, establishing it as the most competitive lightweight baseline. Class-wise analysis indicated Healthy was recognized most reliably, while WhiteFly remained the most challenging due to limited samples, subtle cues, and overlap with healthy leaves. Grad-CAM visualization confirmed the best model focused on symptom-relevant regions rather than background. Overall, findings provide a benchmark for TLID-based classification, identifying EfficientNet-B0 as the strongest lightweight baseline and ResNet18 as the top-performing reference model.},
keywords = {greenhouse tomato leaf disease classification, lightweight convolutional neural network, TLID, EfficientNet-B0, ResNet18, Grad-CAM},
issn = {3069-3187},
publisher = {Institute of Central Computation and Knowledge}
}
Article Metrics
Publisher's Note
ICCK stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and Permissions
Copyright © 2026 by the Author(s). Published by Institute of Central Computation and Knowledge. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
Scan to read this article