To ensure a higher quality, capacity and production of rice, it is critical to diagnose disease early in order to decrease the usage of pesticides and reduce agricultural and environmental damage. Therefore, a Multi-scale YOLO v5 detection network is proposed to resolve rice crop disease in its early stage. The experiment initially starts with the rice leaf images from the RLD dataset for pre-processing, after which data set labels are created, which are then divided into train and test sets. DenseNet-201 is used for the backbone network and depth aware instance segmentation is used to segment the different regions of rice leaf. At next, the proposed Bidirectional Feature Attention Pyramid Network (Bi-FAPN) is used for extracting the features from the segmented image and enhances the detection of disease with different scales. The feature maps are finally identified in the detection head, where the anchor boxes are then applied to the output feature maps to produce the final output vectors by the YOLO v5 network. The experiments are conducted with RLD dataset with different existing networks. The effectiveness of the network is evaluated based on various parameters in terms of average precision, accuracy, average recall, IoU and F1 score.

No abstract