Identifying crop diseases using attention embedded MobileNet-V2 model

Various crop diseases are a major problem worldwide since their occurrence leads to a significant decrease in crop production. The image-based automatic identification of crop diseases that involves food security has attracted much attention recently. It is a challenging research topic due to the complexity of crop disease images, such as clutter field backdrops and irregular lighting strengths. A variety of deep learning networks, especially CNNs, are becoming the mainstream methods for addressing many challenges correlated with image recognition and classification. In this study, to improve the learning ability of minor lesion features, we introduced the Location-wise Soft Attention mechanism to the pre-trained MobileNet-V2, in which the general knowledge of images learned from ImageNet was migrated to our crop disease recognition mode, namely, CDRM. Further, a localization strategy was embedded in the proposed network, and the two-phase progressive strategy was executed for model training. The proposed method shows substantial efficacy in the experimental analyses. It reached a 99.71% average accuracy on the open-source dataset, and even under cluttered background conditions, the average accuracy attained 99.13% for the identification of crop diseases. Experimental findings deliver a competitive performance compared to other state-of-the-art methods and also indicate the efficacy and extension of the proposed method. Our code is available at https://github.com/xtu502/crop-disease-recognition-model. (C) 2021 Published by Elsevier B.V.

Automated summary

Crop diseases are a major issue globally, leading to decreased crop production. Image-based automatic identification methods have gained attention for addressing this problem. This study introduces a Location-wise Soft Attention mechanism in MobileNet-V2, showing promising results in crop disease recognition through experimental analyses.