MOBILE APPLICATION FOR DETECTING SOME WHEAT PATHOGENS USING AI. PUBLISHEDRaluca INCICAȘ*, Otilia COTUNA, F. SALA* None firstname.lastname@example.org
The present study aimed to develop an application for the recognition of five pathogens in wheat culture based on artificial intelligence (AI). Machine learning (ML), an important branch of AI, was the basis for the application of the pathogens recognition in wheat culture. Five pathogens in wheat cultivation were studied, Blumeria graminis, Pyrenophora tritici repentis, Puccinia recondita, Puccinia striiformis and Puccinia graminis. A data set of 323 images with pathogens studied in wheat culture was used. The images have been processed and transformed so that the model receives the same size for each image. The first step in building the data set to train the ML model was data augmentation, in order to increase the number of data through known changes. A training set and a validation set were used. Google Colaboratory was used to build the ML model. The React Native framework was used to have an application available on both iOS and Android. Heroku and Flask were used to integrate the systems. In order to evaluate the model the ”Class Activation Map visualization” (CAM) was used. One of the techniques that CAM uses is to produce heatmaps (areas of interest) on the parts of the image that correspond to the different classes, over the input images. The class activation heatmap is a 2D network of scores associated with a specific output class, calculated for each area of the input image, indicating how important each area is relative to the output class. To view these heatmaps on the images taken in the study, the ClassificationInterpretation class was used. Another method used in evaluating the model was the confusion matrix which will show for each label how many times it was correctly predicted. The model correctly predicted in a percentage of 88.4% for Puccinia striiformis, 72.03% for Puccinia recondita and 94.67% for Blumeria graminis.
artificial intelligence, identification, machine learning, pathogens, wheat
Field crops and pastures