High-dimensional image target defect detection model based on axial self-attention

The invention relates to a target defect detection model based on a small sample high-dimensional image. By introducing an axial self-attention mechanism, a target defect detection model based on a small sample high-dimensional image is designed for solving the problems that a traditional deep convolutional neural network is large in occupied memory, slow in processing time and incapable of effectively solving the target detection problem of a high-dimensional high-definition image in a professional field during image feature extraction. According to the model, effective global representation extraction can be carried out on a high-dimensional picture, the calculation complexity is remarkably reduced, the performance of small sample high-dimensional image target detection is improved, and the requirements of practical application problems are met. According to the target defect detection model based on the small sample high-dimensional image, the problems of slow time and poor effect of traditional deep convolution extraction of the depth global representation of the high-dimensional image are solved, and the method can be effectively used for the practical application problems of defect target detection and the like of the high-dimensional image in a small sample situation.