Detection and counting method of silkworms based on space-to-depth conversion

Silkworm density is an indicator that needs to be focused on in the process of sericulture, which is crucial to the feeding and healthy growth of silkworms. Aiming at the problems of small target of silkworms and blur pictures of silkworms in actual production, this paper proposes a silkworm detection and counting method based on space-to-depth transformation. Firstly, by adding space-to-deep (SPD) modules to the backbone and neck of the YOLOv8 model, the information loss caused by strided convolution and pooling is avoided, which improves the detection ability of small silkworm targets in low-resolution cases; secondly, the detection head of the YOLOv8 model is improved to dynamic detection head, which combines feature information in three different dimensions: scale, space and channel, to enhance the accuracy of silkworm detection and reduce the problem of false detection; finally, the SRN-DeblurNet network is introduced to de-blur the input images, which improves the quality of the image and reduces the effect of image blurring on the detection performance. The experimental results show that the method achieves 87.9%, 90.7.2%, 94.7%, and 79.5% in precision, recall, mAP@0.5 and mAP@0.5:0.95, respectively, which are more obvious advantages compared with similar methods.