首页>
外国专利>
Method and system for glocal description of phytopathology based on Deep learning
Method and system for glocal description of phytopathology based on Deep learning
展开▼
机译:基于深度学习的植物病理学的神经病理学的手工艺系统和系统
展开▼
页面导航
摘要
著录项
相似文献
摘要
The present invention relates to a deep learning-based plant pathology glocal description method, which corresponds to each of the suspicious areas of multiple sizes based on Faster R-CNN (Region-based convolutional neural network) to which FPN (Feature Pyramid Network) is applied. Generating at least one bounding box to be formed, and then obtaining and outputting characteristic information of each bounding box, information on diagnosed symptoms, and context information of the entire image; Based on LSTM (Long-Short Term Memory), it generates and outputs a sentence that specifically describes the location and symptoms of each suspected disease area from the characteristic information of each bounding box and the diagnosed symptom information, and at the same time, from the context information of the entire image. A sentence generation step of generating and outputting a sentence describing an overall situation of the entire image; And calculating a loss index based on accuracy of sentence generation, detection accuracy of a suspected disease region, and detection accuracy of a disease, and end-to-end training of the Faster R-CNN and the LSTM so that the loss index is minimized.
展开▼