Tobacco Leaf Disease Segmentation based on TDSSNet

In recent years, many researchers have utilized deep learning methods to detect and identify plant diseases to help farmers improve the quality of their crops. Tobacco is one of the most important crops. There are few researches on disease identification using deep learning. Most of the existing methods are based on image classification, which can not identify and locate the specific type and location of the lesion. Since the correlation of tobacco diseases is less and there is no relevant dataset, this paper also created tobacco leaf disease dataset on the basis of Tobacco Disease Semantic Segmentation Network(TDSSNet) to study the classification and location of tobacco leaf disease. TDSSNet incorporates Convolutional Block Attention Module(CBAM) and skip connections to enhance feature extraction and fusion. The proposed network achieves a mean Intersection over Union (mIoU) of 64.99% for tobacco leaf disease segmentation.