Concrete slump detection based on light-AWGAM-YOLOv8n.

The precise detection of concrete mortar slump is a key technical link in intelligent concrete mixing, playing a vital role in ensuring project quality and improving construction efficiency. In this paper, the YOLOv8n algorithm is improved to achieve detection of concrete mortar slump efficiently and precisely. At the algorithm level, the AWGAM (Add Weight Global Attention Mechanism) is integrated with the C2f module, and its innovative design has two versions. One is the Basic version B-AWGAM-C2f (Basic-Add Weight Global Attention Mechanism-C2f). The other is A-AWGAM-C2f (Adaptive-add Weight Global Attention Mechanism-C2f). And the two versions are respectively deployed in the neck network and the backbone network of the model to enhance the multi-scale feature fusion ability of the model. Meanwhile, a depth-separable convolution is introduced into the AWGAM attention mechanism to construct a lightweight module, Light-AWGAM. Then it is placed in the backbone network of the model. The number of parameters of the model has significantly decreased while ensuring detection accuracy. The detection precision and the computational efficiency are balanced effectively. Experiments show that the performance of the improved YOLOv8n model has been significantly enhanced: the precision has increased by 3.3%, the recall has increased by 0.6%, and both mAP50 and mAP50-95 have grown by 0.8%. The model demonstrates significant advantages in detection efficiency and accuracy, enabling concrete mortar slump detection tasks to be completed efficiently.