Table 1 Literature review.
Ref. no. | Technique used | Pros | Cons | Accuracy (%) |
|---|---|---|---|---|
Hybrid deep learning model | Adapts well to varied climates, improved disease detection | Computationally expensive | 97.8 | |
Machine learning algorithms | Lightweight and interpretable | Lower accuracy compared to deep learning models | 96.3 | |
Total variation filter-based variational mode decomposition | Enhances image quality for better feature extraction | Computationally intensive preprocessing | 98.8 | |
Lightweight CNN model | Fast inference, suitable for real-time applications | Reduced feature representation capability | 99.3 | |
Hybrid CNN model | Effective for bacterial, viral, and fungal disease classification | Requires large dataset for robust performance | 98.3 | |
Deep learning & machine learning ensemble | High accuracy for plant species classification | Increased training time | 99 | |
Real field dataset with deep learning models | Improved generalization in real-world settings | Susceptible to image noise and occlusions | 90 | |
T-Net lightweight CNN | Efficient for mobile deployment | Lower accuracy on complex disease cases | 98.9 | |
Adaptive ensemble model with exponential moving average fusion | Better feature fusion, enhanced gradient optimization | High computational complexity | 98.7 | |
AI techniques for precision agriculture | Broad applicability in smart farming | Lacks specific model optimization details | N/A | |
Deep learning with explicit control of hidden classes | Reduces misclassification errors | Increased training complexity | 93.3 | |
Comparison of pre-trained CNN models | Identifies the best-performing architecture | Requires large labeled datasets | N/A | |
MobileNetV2 + SVM hybrid model | Efficient feature extraction with enhanced classification | Dependent on dataset quality | 97 | |
Novel feature extraction model for low-computation devices | Optimized for edge computing | Potential accuracy drop on high-resolution images | 87 |
