Lightweight vision transformer and ResNet-9 models for real-time plant disease detection and pest classification with SHAP explainability
Published in Bioengineering & Biotechnology, Earth & Environment, and Sustainability
Context
The rapid advancement of digital technologies alongside increased global commitment to sustainability has intensified the need for efficient crop management solutions. Across agricultural systems, delayed detection and misclassification of plant diseases remain major contributors to reduced yields, threatening food security and undermining progress toward Sustainable Development Goals such as Zero Hunger, No Poverty, Good Health and Well-being, Climate Action, and Life on Land. Plant pests, diseases, and excessive chemical use further exacerbate these challenges. Early, automated visual detection offers a pathway to environmentally responsible and economically viable agricultural practices.
Objective
This study aims to develop and evaluate an automated, real-time plant disease classification framework using Vision Transformers (ViT) and hybrid ViT–CNN architectures, with the goal of supporting farmers and agronomists in early decision-making and sustainable crop protection.
Methods
The research employs deep learning techniques, including ViT and a combined ViT–CNN model built on ResNet-9, trained and evaluated using four publicly available datasets: the Turkey Plant Pests and Diseases (TPPD) dataset (15 classes), the Namibia Maize Image Dataset (3 classes), the Banana Image Dataset (3 classes), and the Tanzania Maize Dataset (3 classes). SHapley Additive exPlanations (SHAP) were applied to generate saliency maps for interpretability. Comparative analyses assessed performance, accuracy, and classification speed across attention-based and hybrid architectures.
Results and conclusions
The proposed model achieved strong performance with 97.4% accuracy, 96.4% precision, 97.09% recall, a 95.7% F1-score, and high agreement measured by Cohen’s Kappa, outperforming existing benchmark models. SHAP visualizations highlighted that the model leverages high-activation areas, edge features, color patterns, texture, shape, and contextual cues in its predictions. While attention-based models improved accuracy, they also caused reduced classification speed. However, integrating attention blocks with CNN layers effectively compensated for this slowdown, achieving both high accuracy and efficient inference. To evaluate the interpretability and deployment feasibility of the proposed model, we considered several parameters are considered to have an integration of faithfulness, localization quality, sparsity, latency, and energy of the models, including pointing accuracy, localization IoU, Centroid Localization Error (pixels), Attribution Sparsity (%), Insertion AUC, Deletion AUC, Time per Explanation (ms/image), Energy Consumption (J/image), Memory Footprint (MB).
Significance
This research provides a transparent and high-performing deep learning solution for plant disease classification, promoting sustainable agricultural development. By reducing reliance on excessive pesticide and herbicide use, enhancing early diagnosis, and improving decision-making, the model supports environmentally responsible farming practices and contributes to global efforts toward food security and ecological resilience. The significance of this evaluation is that it comprehensively assesses the model’s interpretability and real-world deployability by measuring explanation reliability (faithfulness), spatial precision (localization quality), efficiency (latency and memory), and sustainability (energy consumption), ensuring the model is not only accurate but also transparent, efficient, and practical for deployment. Details at: https://link.springer.com/article/10.1186/s12870-026-08667-8
Wasswa Shafik, born in Makindye, Kampala City, Uganda, is a Computer Scientist, Information Technologist, Founder, and Research Director at the Dig Connectivity Research Laboratory (DCRLab), Kampala, Uganda. He received his Bachelor’s in information technology engineering with a minor in mathematics at Ndejje University, Luweero, Uganda, and a Master’s in information technology engineering (communication and computer networks option) at Yazd University, Yazd, Islamic Republic of Iran. He pursued his PhD in digital science (computer science) at the School of Digital Science, Universiti Brunei Darussalam, Brunei Darussalam. His research broadly examines, integrates, and focuses on developing computationally and statistically efficient models and algorithms to address complex questions about artificial intelligence and machine learning problems. His specific research interests include sustainable development, smart agriculture, computer vision, ecological informatics, applied artificial intelligence, the Internet of Things, cybersecurity and privacy, and smart computing. He has authored, coauthored, edited, coedited, and published hundreds of peer-reviewed books, technical papers, and journal papers, and numerous IEEE international conference papers. His “Hirsch index” is 20+ (source: Scopus Author ID 57210203704). He is the coeditor of Blockchain Technologies for Smart Circular Economy and Organisational Sustainability and Transforming Healthcare Sector Through Artificial Intelligence and Environmental Sustainability (Springer Nature, 2025 and 2024, respectively). He is a member of IEEE and served as a reviewer of several international journals indexed in Scopus, Compendex (Elsevier Engineering Index), and WoS He served in different capacities as department support for mathematics for data science, physical computing for intelligence systems, advanced topics in computing, advanced algorithms, and system performance and evaluation. Prior to this, as a department fellow, he served as a research associate at the Intelligent Network Laboratory in Iran. He served in different capacities as a Community Data Officer at the Programme for Accessible Health, Communication and Education; Research Associate and Data Manager at Population Services International; Data Manager and Research Assistant at the Socio-economic Data Center; Research Lead at TechnoServe; and Ag. Chief Executive Officer at Asmaah Charity Organisation.
Follow the Topic
-
BMC Plant Biology
This is an open access, peer-reviewed journal that considers articles on all aspects of plant biology, including molecular, cellular, tissue, organ and whole organism research.
What are SDG Topics?
An introduction to Sustainable Development Goals (SDGs) Topics and their role in highlighting sustainable development research.
Continue reading announcementRelated Collections
With Collections, you can get published faster and increase your visibility.
Leaf senescence and molecular breeding
Leaf senescence is a complex biological process that involves the degradation of chlorophylls and the eventual death of leaf tissues, affecting plant growth and yield. As the final, postmitotic stage of plant life, leaf senescence is a crucial functional shift from nutrient assimilation to remobilization, critical for plant survival. Its timing is regulated by age, hormones, and environmental stress. Consequently, researchers are intensively devising strategies based on known regulatory mechanisms to manipulate initiation and progression of leaf senescence for enhancing crop resilience and productivity.
While many Senescence-Associated Genes (SAGs) have been identified, fundamental knowledge gaps persist, such as: when exactly is leaf senescence initiated? How are signals transmitted between organelles, cells, and tissues? How can we fully map the molecular pathways of cell senescence? Recent advances in genomics and molecular biology have revealed complex regulatory networks that govern this process, enabling to investigate the interactions between molecular, genetic, epigenetic, and environmental factors. The integration of single-cell, multi-omics and CRISPR/Cas9 approaches now enables deeper mechanistic exploration and more precise manipulation of senescence pathways.
At the same time, molecular breeding for optimized or ‘stay-green’ traits is rapidly accelerating, driven by the global demand for climate-resilient and high-yielding crops in support of the United Nations’ SDG 2 (Zero Hunger). Combining mechanistic insights with genome engineering, precision editing, pan-genome–informed allele mining, high-throughput phenotyping, and genomic selection has opened new opportunities to design crops with fine-tuned senescence timing, improved nutrient-use efficiency, and enhanced stress tolerance and photosynthetic productivity. In this context, BMC Plant Biology invites submissions to the Collection Leaf senescence and molecular breeding, aiming to gather research that advances our understanding of senescence mechanisms and translates this knowledge into crop improvement. We invite researchers and experts in the field to submit research articles that explore, but are not limited to, the following topics:
- Multilayered regulation of leaf senescence
- Genetic and epigenetic control of leaf senescence
- Post-transcriptional and post-translational regulation of leaf senescence
- Hormonal regulation of leaf senescence: regulation by classical and peptide hormones
- Regulatory functions and mechanisms of non-hormonal small molecule signals (e.g. NO, H2S) in leaf senescence
- New mechanisms for stress-induced leaf senescence
- Role of environmental factors on the regulation of leaf senescence
- Role of reactive oxygen species in leaf senescence
- Plant metabolic changes during leaf senescence
- Subcellular structural dynamics and functional remodelling in chloroplast and mitochondria during leaf senescence
- Novel signaling components in regulating leaf senescence in crops
- Integration of leaf senescence traits in molecular breeding programmes
- Pan-genome, GWAS, and QTL analyses for senescence traits
- Marker-assisted selection, genomic selection, and genome editing for senescence improvement
- High-throughput and image-based phenotyping of senescence in crop breeding
All manuscripts submitted to this journal, including those submitted to collections and special issues, are assessed in line with our editorial policies and the journal’s peer review process. Reviewers and editors are required to declare competing interests and can be excluded from the peer review process if a competing interest exists.
Publishing Model: Open Access
Deadline: Aug 19, 2026
Omics technologies and applications in horticultural crops - Part II
The rapid advancement of high-throughput omics technologies has enhanced our understanding of the molecular mechanisms that regulate plant growth, development, and stress adaptation in horticultural crops. Omics technologies and their applications enable the identification of genetic and molecular mechanisms underlying key traits and regulatory responses in horticultural crops, thereby allowing targeted breeding efforts towards horticultural varieties and germplasm with enhanced agronomic traits, nutritional quality and stress resilience. In particular, the integration of omics approaches for instance with systems biology, synthetic biology, computational biology and artificial intelligence-based technologies, can provide a comprehensive framework for elucidating gene functions and regulatory networks controlling important traits such as crop yield, fruit ripening, and stress tolerance. These advances are crucial to develop novel breeding strategies, and optimize crop management and sustainable agricultural practices, particularly under changing climate conditions.
In support of the United Nations’ Sustainable Development Goals SDG 2 (Zero Hunger) and SDG 13 (Climate Action), BMC Plant Biology presents the Collection Omics technologies and applications in horticultural crops - Part II. This Collection focuses on the development and applications of omics technologies in horticultural crops, seeking to highlight recent advances in (meta-)genomics, (meta-)transcriptomics, proteomics, metabolomics, phenomics, epigenomics and ionomics in horticultural crop research. We highlight research articles that explore, but are not limited to, the following topics:
- Application and integration of omics technologies (e.g. genomics, meta-genomics, transcriptomics, metabolomics, fluxomics, proteomics, epigenomics, phenomics) for horticultural crop improvement of yield, pathway dissection, stress resilience, nutrition, and other quality traits
- Multi-omics approaches for horticultural crop breeding and germplasm improvement
- Functional characterization of genes and regulatory networks in horticultural crops using omics tools
- Genome-wide association studies and genomic selection for morphological and agronomic traits, as well as biotic and abiotic stress tolerance in horticultural crops
- Omics technologies and approaches for disease and abiotic stress resilience in horticultural crops
- Omics technologies and approaches in sustainable agriculture and precision agriculture
- Multi-omics approaches applied to plant-pathogen interactions in horticultural crops
- Genomics-assisted breeding of horticultural crop varieties for yield improvement and disease tolerance
- Single-cell and spatial omics applied to horticultural crop research
- Long-read genomics and epigenomics applied to horticultural crops
- Small RNA-omics to study regulatory networks associated with horticultural crop traits
- Multi-omics approaches for improving fruit quality, ripening and postharvest biology, and source-sink regulation in horticultural crops
- Plant phenomics and ionomics for trait characterization, nutrient use efficiency, and yield improvement in horticultural crops
All manuscripts submitted to this journal, including those submitted to collections and special issues, are assessed in line with our editorial policies and the journal’s peer-review process. Reviewers and editors are required to declare competing interests and can be excluded from the peer review process if a competing interest exists.
Publishing Model: Open Access
Deadline: Dec 28, 2026
Please sign in or register for FREE
If you are a registered user on Research Communities by Springer Nature, please sign in