Attention-Concatenation Dense Convolutional Neural Network for Silkworm Disease Recognition

Silkworm breeding is an important economic source in Guangxi, China, but silkworm disease has a huge impact on silkworm disease. If silkworm disease cannot be identified accurately, it will cause economic losses to silkworm farmers. With the maturity of machine learning theory and the development of image processing technology, it is possible to recognize the silkworm disease image by the algorithm of CNN. In order to realize the accurate recognition of silkworm disease, This paper is based on DenseNet, combining with the characteristics of silkworm disease images, the information of different levels is different and the details are rich, by improving the model structure, an image retrieval algorithm for silkworm disease based on attention mechanism was proposed(AC-DenseNet). By setting the contrast experiments, the results of different CNN algorithms are compared and analyzed. 600 images of silkworm disease were used in the experiment, the recognition rate of AC-DenseNet reached 85.6%, which was 8.9% and 13.9% higher than the comparison algorithms SENet and DenseNet, respectively, which verified that the AC-DenseNet proposed in this paper has a good recognition rate for silkworm disease image recognition.