The McKinsey Way: Using The Techniques Of The W...
A high-throughput phenotyping method was developed to assess soft rot disease symptoms using Python and MATLAB image analysis software, using neural networks that successfully extract the exact lesions that are in each inoculated tuber. Semantic segmentation based on deep learning was used in the process of disease isolation. The U-net neural network was adapted, which was appropriately modified by adding the normalization and early stop functions. A total of 2700 images were used to train the neural network, which were generated on the basis of 400 previously labeled images using data augmentation techniques such as blurring, rotation, and image sharpening, which were intended to protect the training of the network against overtraining. A high accuracy value of 0.95 was obtained for the identification of diseased areas, which were used for further statistical analysis [96].
The McKinsey Way: Using the Techniques of the W...
041b061a72