Light-weight bridge apparent disease detection method based on deep learning

The invention discloses a light-weight bridge apparent disease detection method based on deep learning, and relates to the technical field of concrete building safety assessment. According to the method, a small target detection layer network is improved in the YOLOv8 neural network, a large target detection layer of an original detection network is abandoned, meanwhile, LMConv convolution is introduced into a C2f layer of the YOLOv8 neural network, the parameter quantity is reduced, meanwhile, the disease feature extraction capacity is enhanced, and precision is improved; respectively combining two convolutions on the BBox Loss branch and the Cls Loss branch on the YOLOv8 detection head, replacing the two convolutions with a group convolution, and carrying out parallel calculation to reduce the calculation amount; and meanwhile, Inner-IoU is introduced, limitation of an original loss function is broken through by using an auxiliary bounding box, convergence of the network is accelerated, and the detection ca