【摘要】 针对钢材表面缺陷检测中存在检测精度低、模型尺寸大等问题，提出一种基于YOLOv7-tiny网络改进的算法模型PC-YOLOv7。首先将PC-ELAN结构替换主干网络中部分ELAN结构，降低模型参数量和模型尺寸；其次在特征融合网络(Neck)部分采用双向特征金字塔网络(bidirectional feature pyramid network, BiFPN)增强图像高层语义信息和低层特征信息融合性能，在输出端引入SPD-Conv提高模型对低分辨率物体的检测能力；最后，提出SimCS-CA模块并引入特征融合网络增强模型的特征表示性能。实验结果表明，PC-YOLOv7算法在NEU-DET数据集上平均精度均值(mAP)达到了78.5%,相比原始YOLOv7-tiny算法在模型尺寸降低情况下准确率和mAP分别提升了10.6%和4.2%,验证了改进算法的有效性。更多还原

【Abstract】 In order to solve the problems of low detection accuracy and large model size in steel surface defect detection, an improved PC-YOLOv7 algorithm model based on YOLOv7-tiny network is proposed. First, the PC-ELAN structure is used to replace part of the ELAN structure in the backbone network to reduce the number of model parameters and model size. Secondly, in the Neck part, bidirectional feature pyramid network(BiFPN) is used to enhance the fusion performance of high-level semantic information and low-level feature information of the image. The SPD-Conv layer is introduced at the output to improve the model′s ability to detect low-resolution objects. Finally, the SimCS-CA module is proposed and the feature fusion network is introduced to enhance the model′s feature representation performance. Experimental results show that the mean average precision(mAP) of the PC-YOLOv7 algorithm on the NEU-DET dataset is 78.5%. Compared with the original YOLOv7-tiny algorithm, the accuracy and mAP are improved by 10.6% and 4.2% respectively when the model size is reduced, which verifies the effectiveness of the improved algorithm.更多还原