Identification of Maize Seed Varieties Using MobileNetV2 with Improved Attention Mechanism CBAM

Seeds are the most fundamental and significant production tool in agriculture. They play a critical role boosting output revenue of To achieve rapid identification protection maize seeds, 3938 images 11 different types seeds were collected for experiment, along with combination germ non-germ surface datasets. The training set, validation test set randomly divided by ratio 7:2:1. experiment introduced CBAM (Convolutional Block Attention Module) attention mechanism into MobileNetV2, improving replacing cascade connection parallel connection, thus building an advanced mixed module, I_CBAM, establishing new model, I_CBAM_MobileNetV2. proposed I_CBAM_MobileNetV2 achieved accuracy 98.21%, which was 4.88% higher than that MobileNetV2. Compared to Xception, MobileNetV3, DenseNet121, E-AlexNet, ResNet50, increased 9.24%, 6.42%, 3.85%, 3.59%, 2.57%, respectively. Gradient-Weighted Class Activation Mapping (Grad-CAM) network visualization demonstrates focuses more on distinguishing features seed images, thereby model. Furthermore, model is only 25.1 MB, making it suitable portable deployment mobile terminals. This study provides effective strategies experimental methods identifying varieties using deep learning technology. research technical assistance non-destructive detection automatic varieties.