Table 1.
Summary of related studies. The accuracy values in the last column correspond to the datasets listed in the study and in the order they are given in the previous column
| Ref | Models | Application | Dataset | Performance (accuracy %) |
|---|---|---|---|---|
| [21] | NB, DT, KNN, SVM, RF | Disease classification in maize | PlantVillage dataset | 79.23 % (Random Forest) |
| [22] | CNN-based method | Plant disease detection | PlantVillage | 88.80% |
| [23] | Pre-trained models (VGG, ResNet, DenseNet) | Plant disease classification | Plant Village | 98.27% (DenseNet) |
| [24] | ResNet50, Xception, MobileNet, ShuffleNet, Densenet121_Xception | Tomato leaf disease classification | PlantVillage | 97.10% (Densenet_Xception) |
| [25] | VGG16 | Tomato plant disease classification | Plant Village | 95.50% |
| [26] | EfficientNet-b0 through EfficientNet-b7, EfficientNetv2-small, EfficientNetv2-medium, EfficientNetv2-large, ResNetv2-50, and InceptionV4 | Disease detection in sugarcane leaves | Sugarcane Leaf Dataset | EfficientNet-b6 (93.39%) |
| [28] | 14 CNN and 17 vision transformer models | Classification of grape leaves and diagnosis of grape diseases | PlantVillage and Grapevine datasets | CNN + ViT (Swinv2-Base) (100%) |
| [27] | ResNet50, InceptionV4, Xception, DenseNet121, EfficientNetV2_m, and VGG13 | Classification of apple diseases (on leaves) | PlantVillage dataset | EfficientNetV2_m (100%) |
| [16] | Res2 Next50, Res2 Net50 d, VGG16, and DenseNet121 | Detecting diseases in tomato leaves | Small dataset with 13,875 tomato images | Res2 Next50 (99.85%) |
| [29] | ViT, hybrid of CNN and ViT | Real-time automated plant disease classification | Wheat Rust, Rice Leaf Disease and Plant Village | Balance between accuracy and prediction speed |
| [30] | ViT (GreenViT) | Plant disease detection | Plant Village, Data Repository of Leaf Images and a merged dataset | 100.00%, 98.00% and 99.00% respectively |
| [31] | ViT (FormerLeaf) | Cassava leaf disease detection | Cassava leaf disease dataset | Reduce model size by 28.00% and decrease inference speed by 10.00% |
| [32] | Hybrid model (ViT + CNN) | Plant disease detection | Plant Village and Embrapa | Accuracy of 98.86% and 89.24% respectively |
| [33] | ViT (PMVT) | Real-time detection of plant diseases | wheat, coffee, and rice | 93.60%, 85.40% and 93.10% respectively |
| [34] | Inception Convolutional ViT | Automatic plant disease identification | PlantDoc, AI2018, PlantVillage, ibean | 77.54%, 86.89%, 99.94%, and 99.22% |
| [35] | ViT enabled CNN (PlantXViT) | Plant disease identification | Apple, Embrapa, Maize, PlantVillage, and Rice | 93.55%, 89.24%, 92.59%, 98.86%, and 98.33% |
| [36] | Image processing and machine learning-based system | Potato leaf disease identification and classification | PlantVillage | 97.00% (Random Forest) |
| [14] | PLDPNet (VGG19 + Inception-V3 + ViT) | Potato leaf disease classification | Plant Village | 98.66% accuracy, 96.33% F1-score |
| [37] | EfficientRMT-Net (ViT + ResNet50) | Potato leaf disease classification | Plant Village (General, specialized) | 97.65% (general), 99.12%(specialized) |
| [38] | InceptionV3, VGG16 and VGG19 | Potato leaf disease detection | Plant Village | 97.80% (VGG19 + logistic regression) |
| [13] | DenseNet201 | Potato leaf disease classification | Plant Village, additional data | 97.20% |
| [39] | VGG 16, VGG 19, MobileNet and ResNet50 | Late and early blight diseases recognition in potato crops | Plant Village | 97.89% (VGG 16 after fine-tuning) |