Agricultural insect pest image detection method based on C3M-Yolov5s

The invention provides an agricultural insect pest image detection method based on C3M-Yov5s, and the method comprises the steps: 1, replacing a standard convolution Conv module of an original model Yov5 with a convolution module provided in MobileNetV3 in order to reduce the parameter quantity and calculation cost of the original model while carrying out the convolution extraction of image features, i.e., accelerating the convergence speed of a network; secondly, a bottleneck type convolution module of MBottleneck is built on the basis of a Bottleneck structure, the structure is more flexible when image features are extracted under the condition of guaranteeing a relatively low parameter quantity, and features extracted after the size of a convolution kernel is scaled have higher specificity; thirdly, a C3M is built by using MConv and Mmottleneck for reference to a stagger structure, the stagger structure can effectively avoid the defect of gradient disappearance, the number of layers of the network is reaso