CropCapsNet: Enhanced capsule network for crop disease classification

Abstract

The prevention and treatment of crop diseases are crucial for the development of smart agriculture. The classification of crop diseases based on deep learning for early disease monitoring and control has become the mainstream direction of research. This paper proposes a novel deep learning model called ”CropCapsNet”, which combines Squeeze-and-Excitation Inception (SE-Inception) module and has improved capsule structure for crop disease classification. The network first extracts shallow features of input samples through double-layer convolution, then uses SE-Inception to achieve deep multi-scale feature acquisition, and finally outputs classification results through an improved capsule structure. SE-Inception adds Squeeze-and-Excitation(SE) attention after each multi-scale feature extraction block to improve the model’s perception of diseases without increasing the number of parameters. The improved capsule structure is embedded with a parameter grouping strategy, which can control trainable parameters by adjusting the number of capsule groups to adapt to different application scenarios. To verify the generalization of the network, this paper uses three datasets containing different experimental scenarios (PlantVillage, Xinong Apple Dataset, and FGVC8) to evaluate the performance of CropCapsNet. The results show that CropCapsNet has achieved classification accuracies of 99.99%, 98.18%, and 98.09% in the three datasets, respectively. Compared with methods such as ConvNeXt, RegNet, and ResNeSt, CropCapsNet performs excellently. In addition, this paper uses image reconstruction networks and heatmaps to visualize CropCapsNet, improving the interpretability of the model.