[24] | 2020 | Single | Grape | leaf | Leaf Blight | early, middle & end | Plant Village (1293 images) | AlexNet and ResNet18 | AlexNet accuracy: 90.31%; ResNet accuracy: 87.6% |
[4] | 2021 | multi-task | Pear | Leaf | Biotic stresses: leaf spot, leaf curl, and slug damage | No risk (0%), very low (1% - 5%), low (6% - 20%), medium (21% - 25%), and high (>50%) | DiaMOS Plant dataset, a self-collected dataset | ResNet50, VGG-16, VGG-19, MobileNetV2, EfficientNetB0 and InceptionV3 | InceptionV3 obtained bests accuracies of 90.68% and 74.07% for biotic stress and severity estimation, respectively. |
[44] | 2021 | single | * | * | * | * | Self-collected dataset and Plant Village dataset | Proposed lightweight CNN | accuracies of 97.9% and 90.6% on the Plant Village dataset and plant disease severity dataset, respectively |
[21] | 2021 | single | Patato | leaf | late blight lesion | * | Self-collected dataset of 70 original images | Proposed Deep Learning algorithm | IoU values of background (soil and leaf) and lesion classes in the test dataset are 0.996 and 0.386, respectively. |
[23] | 2021 | Single | Wheat | Leaf | Yellow rust | No disease, resistant, moderately resistant, moderately susceptible, or susceptible | Self-collected dataset of 10,500 images | Yellow-Rust-Xception | Accuracy: 91% |
[11] | 2021 | Single | Cucumber | leaf | Angular Spot, Anthracnose, Black Spot, Brown Spot, Downy Mildew, Gray Mold, Powdery Mildew and Target Spot. | percentage | Plant Village (689 images) | proposed CNN | Accuracy: 93.75% |
[38] | 2021 | Single | Tomato | leaf | Spotted Wilt | early, middle and least | Self-collected (3000 images) | proposed CNN | binary classification: 91.56% of accuracy and multi-classification: 95.23% of accuracy |
[35] | 2021 | Multi-task | mustard | leaf | Downy mildew | 4 severity levels | Self-collected | proposed CNN | Binary classification: 95.6% of accuracy and multi-classification: 96.66% of accuracy |