Research on ResNet50 forage image classification algorithm integrated with SE attention

Forage classification is of great significance in the field of agriculture and animal husbandry, and it is important for livestock feeding, forage species selection, and grassland protection. In this paper, the model based on deep learning is proposed to improve the accuracy and efficiency of pasture image classification. Firstly, the pasture image dataset is constructed to complete preprocessing, including image enhancement, resizing, and labeling. Secondly, the ResNet50 model under PyTorch is adopted as the image classification model to improve the model generalization ability, and the SE attention mechanism is added to mix the depth separable convolution. Finally, the model parameters are optimized by cross-entropy loss function and stochastic gradient descent algorithm. In this paper, a total of 13 species of forage grasses were collected, such as ice plant, awnless birdseed, alfalfa, pigweed, and oxalis. The accuracy rate reaches 99.44%. The experimental results indicate the method has achieved significant improvement in accuracy and efficiency, proving the effectiveness and feasibility of ResNet50 in forage image classification.