SE-Resnet152 Model: Early Corn Leaf Disease Identification and Classification using Feature Based Transfer Learning Technique

Various corn leaf diseases reduce the quantity and quality of corn crop production, so early detection and classification are important for preventing crop yield. However, the detection and classification of corn leaf disease are more difficult due to the regions of the leaf blur and noise effect. To solve the above-mentioned problems, a feature-based transfer learning approach called the Convolutional Deep Learning (CDL) model is proposed. First, the images of corn leaves taken from Plant Village are preprocessed with Green Channel Conversion (GCC) and Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithms to enhance low-contrast images. Then the important features are extracted through ResNet152, which contains two fully connected layers with a sigmoid function. Next, squeeze and excitation are combined with ResNet152 to gain better performance in detection and classification. The proposed SE-ResNet152 model outperforms with a Precision 98.71%, Recall 98.14%, F1-score 96.87% and accuracy 99.02% respectively compared to existing model like U-Net and ResNet50 models.