Classification and identification of crop disease based on depthwise separable group convolution and feature fusion

Crop diseases have a significant impact on crop yield and quality, so there is an urgent need for a fast and accurate detection method to identify the types of diseases and treat them in a timely manner. In this study, we propose an improved method for crop disease detection using the grape, corn, and tomato disease datasets from PlantVillage. By introducing the depthwise separable group convolution (DSGC) technique and the adaptive feature fusion mechanism (AFFM), this method combines with the convolutional neural network (CNN) model to reduce model computational resource consumption while enhancing the model's ability to learn feature information at different scales. Experimental results show that the accuracy, precision, recall, and F1-score of crop disease classification using the Resnet50 model + AFFM are 99.56%, 99.53%, 99.56%, and 99.57%, respectively. This confirms the effectiveness of the proposed method in successfully distinguishing different crop diseases. Therefore, the method proposed in this study provides a new and efficient solution for crop disease detection and has broad application prospects.