Crop Pest and Disease Recognition Based on Improved ConvNeXt and Swin Transformer

Agriculture is essential for global food security, which significantly impacts human happiness and overall wellbeing. As the population grows, improving agricultural productivity becomes crucial for ensuring a stable food supply and supporting societal welfare. Pests have long affected crop yields, threatening both food security and quality of life. This paper leverages deep learning by combining ConvNeXt and Swin Transformer models. A newly proposed attention mechanism, CAFM Fusion, enhances feature fusion, while Mixup and dynamic learning rate adjustment help prevent overfitting. Tested on IP102 and PlantVillage datasets, our model achieved 75.75% and 99.89% accuracy, respectively, contributing to sustainable agriculture and better well-being.