Doodernet: Enhancing Black Gram Yield Through Early Detection Of Cuscuta Infestation
DOI:
https://doi.org/10.64252/bav70890Keywords:
Cuscuta detection, Deep learning segmentation, Parasitic plant identification, Agricultural computer vision, U-Net architecture, Crop disease managementRetryClaude can make mistakes. Please double-check responses.Abstract
Cuscuta, commonly known as dodder, is a parasitic plant that poses a significant threat to crops worldwide, including black gram (Vigna mungo). Early detection and management of Cuscuta infestation are crucial to minimize yield losses and ensure optimal crop productivity. In this paper, the DodderNet is used to detect Cuscuta early in black gram plant farms. The proposed approach, DooderNet, combines preprocessing techniques, Image Annotation, segmentation, and U-Net. The pre-trained model RESNET50 is used to train on given Cuscuta images. It isn't easy to train these images with this model. The dataset consists of 200 high-resolution images of black gram fields in Avanigadda Mandal, Krishna District, Andhra Pradesh, India, and corresponding annotations in jpg image format. Finally, there are comparisons between various state-of-the-art convolutional neural network (CNN) architectures, including U-Net, DeepLab, and Fully Convolutional Network (FCN). The existing DeepLab and FCN, compared with DooderNet, show robust performance with an accuracy of 0.90%.
