DOI: 10.22213/2658-3658-2018-38-44

Article

The Use of Convolutional Neural Network LeNet for Pollen Grains Classification

Korobeynikov, A., Kamalova, Yu., Palabugin, M., Basov, I.

Abstract. The convolutional neural network from the LenetMnistExample of the DeepLearning4j framework is described and applied for pollen grains classification. The selected basic topology of the LeNet neural network was not changed; the loading of images was modified, and the number of classes of the classification task (outputNum) as well as the subsample size of examples (batchSize) were changed. Training (1520  photos) and test (380 photos) samples of four classes of pollen grains were formed. The quality metrics values calculated according to the results of the test sample classification are: 1) Accuracy = 0.9289; 2) Precision = 0.9306; 3) Recall = 0.9266; 4) F1 Score = 0.9282.

Keywords: computer pollen analysis, pollen of plants, image recognition, convolutional neural network, pollen recognition

Pages: 38–44Total pages: 7

Published in: Instrumentation Engineering, Electronics and Telecommunications – 2018: Proceedings of the IV International Forum (December 12–14, 2018, Izhevsk, Russian Federation)

Year of publication: 2018