International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 13 Issue: 01 | Jan 2026
p-ISSN: 2395-0072
www.irjet.net
MS-HAN: A Multi-Scale Hierarchical Attention Network with Multi-Head Architecture for Generalized Crop Disease Detection Aishwarya Malviya1, Surendra Gupta2 1Master of Technology, Dept. of Computer Engineering, SGSITS, Indore, Madhya Pradesh 2Associate Professor, Dept. of Computer Engineering, SGSITS, Indore, Madhya Pradesh
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Plant diseases represent a formidable threat to
shift towards automated, data-driven plant pathology [12], [14]. However, the direct application of generic.
global food security, precipitating annual economic losses estimated to exceed $220 billion worldwide. While deep learning offers a promising avenue for automated diagnostics, existing models often falter when faced with the challenges of scale variance in symptoms, limited generalization across different crop types, and high computational demands. This paper introduces MS-HAN (Multi-Scale Hierarchical Attention Network) with a novel multi-head architecture, meticulously engineered for robust, generalized plant disease detection. MSHAN integrates three principal innovations: (1) an Inceptionstyle multi-scale module that concurrently captures disease features at multiple granularities, (2) a novel hierarchical attention mechanism, analogous to CBAM, that sequentially refines features along channel and spatial dimensions to focus on the most salient pathological indicators, and (3) a multihead classification architecture that separately identifies crop types and disease conditions for improved generalization. Evaluated on a comprehensive, aggregated dataset of over 53,000 images spanning 44 disease classes across 12 crops, MS-HAN achieves a combined test accuracy of 94.78%, with individual crop and disease classification accuracies of 96.2% and 95.4% respectively. It significantly outperforms con temporary models while maintaining a lightweight profile (4.8M parameters), making it suitable for practical edge deployment. Our results underscore the efficacy of a taskspecific, attention driven multi-head design in creating a unified solution for precision agriculture.
Convolutional Neural Network (CNN) architectures confronts three fundamental limitations in this domain: • Inadequate Handling of Scale Variations: Disease symptoms are morphologically diverse, ranging from minute foliar spots to large necrotic lesions. Standard CNNs, with their fixed-size kernels, often fail to capture this wide spectrum of feature scales effectively [16], [21]. • Limited Cross-Crop Generalization: Many existing models are designed and optimized for a single crop type [17], [23]. This specificity severely limits their practical utility in diverse agricultural ecosystems where multiple crops are cultivated [13], [15]. • Single-Head Classification Limitation: Traditional approaches use a single classification head that must simultaneously identify both crop type and disease condition, creating an unnecessarily complex learning problem and limiting generalization across crops with similar dis eases [22]. To overcome these challenges, this research proposes the MultiScale Hierarchical Attention Network (MS-HAN) with a novel multi-head architecture. Our work provides a unified solution for multi-crop disease detection with the following core contributions: • Multi-Scale Processing: An Inception-style module with parallel convolutional branches and varied kernel sizes to capture disease patterns at multiple resolutions. • Hierarchical Attention Mechanism: An integrated attention module performing sequential channel and spatial attention to focus on salient features. • Multi-Head Architecture: Separate classification heads for crop identification and disease detection to improve learning and generalization. • Cross-Crop Generalization: Validated on a large-scale multi-crop dataset (12 crops, 44 diseases). • Computational Efficiency: A lightweight architecture (4.8M parameters) suitable for edge deployment.
Key Words: Plant Disease Detection, Deep Learning, Multi Scale Features, Attention Mechanism, Multi-Head Architecture, Agricultural Computer Vision, Precision Agriculture.
1.INTRODUCTION
2. LITERATURE SURVEY
Plant diseases pose a severe and persistent threat to global food security, with estimated annual losses of $220 billion globally [10]. In India alone, these losses amount to approximately $30 billion annually [11]. The conventional approach to disease management, relying on manual visual inspection, is not only labor-intensive and subjective but also critically constrained by the availability of expert agronomists, rendering it inadequate for the scale and immediacy required in modern agriculture [13]. The advent of deep learning-based solutions has catalyzed a paradigm
© 2026, IRJET
|
Impact Factor value: 8.315
2.1 CNNs for Plant Disease Detection CNNs form the foundation of modern automated plant disease detection systems. Early pioneering work by Mohanty et al. [1] demonstrated the potential of deep learning, achieving an accuracy of 99.35% on the PlantVillage dataset using a pre trained GoogLeNet model. Subsequent studies explored various architectures; Pandian et al. [2] developed a 9-layer custom CNN achieving 97.8% accuracy. However, a
|
ISO 9001:2008 Certified Journal
|
Page 118