Table 1 Comparison of existing studies related to wheat disease detection.
From: Deep learning framework using UAV imagery for multi-disease detection in cereal crops
Study/Author | Method/Model | Crop(s) | Classification Type | Image Complexity | Key Findings/Accuracy | Limitations/Gap |
|---|---|---|---|---|---|---|
Singh, H. et al.33 | Artificial Neural Network (ANN) | Wheat | Binary classification | Satellite images | Detect yellow rust disease using sentinel-2 satellite images on a local scale, 91% accuracy | Limited to rust disease only, Satellite data, limited generalization to real-field images. No possibility of a global model |
Nigam et al.16 | MBConv (EfficientNet B0–B7) | Wheat | Multi-class classification | handheld mobile camera images | Detected wheat rust disease, achieving 99% testing accuracy | Limited to rust disease types only, low multi-crop generalizability |
Tang et al.37 | Pre-trained ResNet-18 | Wheat | Binary classification | Multi-leaf images | Stripe rust detection in diverse fields, 86% accuracy | Limited feature extraction under complex conditions |
Chen et al.38 | INC-VGGN (VGG + Inception + Global Pooling) | Rice | Multi-class classification | single leaf images with plain background | Improved accuracy and feature extraction, 92% accuracy | Complexity, specific disease focus |
Rangarajan Aravind, K. et al.39 | VGG16/19, ResNet101, DenseNet201 | 4 crops | Multi-class classification | single leaf images with plain background | Multi-class classification, VGG16 highest accuracy of 90% in real-field conditions | Limited real-field generalization |
Bukhari, H. R. et al.45 | U² Net segmentation | Wheat | Multi-class classification | uniform background, single leaf images | yellow rust disease detection with 96.19% accuracy | Limited crop types, disease-specific |
Hayıt, T. et al.47 | CNN-CGLCM_HSV + SVM | Wheat | Multi-class classification | single leaf images | Outperformed single feature models, 92.4% accuracy | Limited multi-disease detection and real-field variability |
Present Study | Hybrid fine-tuned VGG16 + PCA + ML classifiers | Wheat | Multi-class classification | Multi-leaf, in-field complex background images | 97% testing accuracy, detects 4 wheat diseases under complex real-field conditions | Addresses multi-disease, UAV-based, real-field, complex imagery, hybrid DL + ML framework |
