Table 3 A comparative study of the proposed architecture compared with previous work for tomato disease recognition
From: An optimized capsule neural networks for tomato leaf disease classification
Researchers | Methodology | Dataset specification | Class labels | Accuracy |
|---|---|---|---|---|
Mokhtar et al., 2015 [45] | Support Vector Machine (SVM) | The 200 infected tomato leaf images dataset includes two different yellow leaf curl diseases | Two Yellow Leaf Diseases | 90.00% |
Brahimi et al., 2017 [47] | Using CNN compared with shallow models and hand-crafted features | The dataset contains 14,828 images for nine diseases | 4032 Yellow Leaf Curl Virus 1. 325 Mosaic Virus 2. 1356 Target Spot 3. 1628 Spider Mites 4. 904 Leaf Mold 5. 1723 Septoria Spot 6. 1781 Late Blight 7. 952 Early Blight 2127 Bacterial Spot | 94.53%, and 95.46% |
Sardogan et al., 2018) [44] | CNN model and Learning Vector Quantization (LVQ) algorithm | The dataset with 400 leaf images for training and 100 images for testing the following diseases | Healthy Bacterial Spot Late Blight Septoria Spot, Yellow Curved | 86.00% |
Zhang et al., 2018 [60] | CNN and the applied transfer learning algorithms for Resnet, Alexnet, and GoogleNet | The dataset contains 5550 images for Health and other eight diseases included | Early Blight, Yellow Leaf Curl Disease, Yellow Leaf Curl Virus, Leaf Spot, Leaf Mold disease, Mosaic Virus, Late Blight, Two-spotted Spider Mite | 96.51% |
Foysal et al., 2020 [46] | CNN | The dataset has 600 input images, 100 for each class, and six class leaf diseases | Healthy Bacterial Spot Late Blight Septoria Spot, Yellow Curved Spider Mites | 76.00% |
Abbas et al., 2021 [48] | Conditional Generative Adversarial Network (C-GAN) with DenseNet121 model | PlantVillage dataset contains ten categories of diseases with 16,012 images | Tomato Yellow Leaf Curl Virus Tomato Bacterial Spot Tomato Late Blight Tomato Septoria leaf spot Tomato Two Spotted Spider Mite Tomato Target Spot Tomato Early Blight Tomato Leaf Mold Tomato Mosaic Virus Tomato healthy | DenseNet, C-GAN 97.11% |
Atila et al., 2021 [49] | EfficientNet | PlantVillage dataset contains 39 plant leaf diseases and 10 class labels of tomato leaf images The number of tested images of tomatoes was 500 leaf images | Tomato Bacterial Spot Tomato Early blight Tomato Late blight Tomato Leaf Mold Tomato Septoria leaf spot Tomato Spider mites Tomato Target Spot Tomato Yellow Leaf Curl Tomato mosaic virus Tomato healthy | Average accuracy 99.00% |
Chowdhury et al., 2021 [50] | EfficientNet and modified U-net | PlantVillage dataset contains 16,485 images with ten class labels | Healthy Early Blight Septoria Leaf Spot Target Spot Leaf Mold Bacterial Spot Late Bright Mold Tomato Yellow Leaf Curl Virus Tomato Mosaic Virus | Accuracy 99.17% |
Tan et al., 2021 [51] | KNN SVM RF AlexNet VGG16 ResNet34 EffeicientNet MobileNetV2 | Ten class labels of PlantVillage dataset with 1591 healthy and 5357 infected tomato images | Bacterial Spot Early Blight Late Blight Leaf mold Septoria leaf spot Two-spotted spider mite Target spot Tomato mosaic virus Tomato yellow leaf curl virus Health | Accuracy KNN = 82.10% SVM = 91.00% RF = 82.70% AlexNet = 92.70% VGG16 = 98.90% ResNet34 = 99.70% EfficientNet = 98.90% MobileNetV2 = 91.20% |
Proposed Architecture | Traditional CNN and Capsule Network-based Adam optimizer | The dataset contains 10 categories; Trained images • 58,122 Tested images • 12,712 The total number of images • 70,834 | Two-spotted Spider Mite, Target Spot, Tomato Mosaic Virus, Yellow Leaf Curl Virus, Bacterial Spot, Early Blight, Late Blight, Leaf Mold, Septoria Leaf Spot Healthy Leaves | Traditional CNN = 92.87% Capsule network accuracy = 96.39% |