A Method for Potato Leaf Disease Recognition Based on Position Encoding and BAM Attention Mechanism
Accurate detection and recognition of potato leaf diseases is essential for precise pest and disease control,and can effectively improve potato yields.However,due to the similarity in early manifestations between early and late diseases in potato leaves,it is difficult to distinguish them.In order to detect and recognize potato leaf diseases more accurately,this paper proposes a ConvNeXt model based on positional encoding and parallel attention mechanism.Firstly,the dataset is preprocessed with position encoding,so that the network model can obtain the position information of disease sites without loading pre-training weights,which improves the learning ability;Secondly,for the different spatial distribution locations of different diseases and the slight differences in morphological features,the BAM module of the parallel attention mechanism is added to enhance the model's ability of extracting disease features.The experimental results show that the optimized ConvNeXt model is able to accurately detect and classify different diseases,with a maximum increase of about 5 percentage points in accuracy compared with the original ConvNeXt model Top-1,which is able to satisfy the current demand for accurate recognition of potato leaf diseases,and it has good robustness,and can be generalized to other plant species.
NETL
NSTL
万方数据
