Identification of Wild Mushroom Based on Ensemble Learning

Yunnan Province has the most abundant wild mushroom resources in China. With the development of modern science and technology, wild mushroom resources have attracted more and more attention because of their potential edible, medicinal, and economic value. The identification of wild mushrooms is regarded as an important research field by more and more researchers. In this paper, we collected the data of wild mushrooms in multiple scenes as the training data of the base learners. The VGG16, Resnet18, and Googlenet models integrated by bagging algorithm are trained. Then they are used for integration learning. So that adapt to the complex scenes of wild mushroom recognition and improve the accuracy and generalization ability. Experiments have shown that, compared with a single CNN, our model showed better performance. With our 10% Holdout Validation Dataset, we saw that the accuracy of the integrated model is 93.1%, compared with the best single bagging integrated model 90.8%.