基于深度学习算法的建筑垃圾分类检测技术对建筑垃圾回收和资源再利用具有重要意义。提出了改进的YOLOv7算法实现对建筑垃圾的分类检测。改进算法采用内容感知特征重组(content-aware reassembly of features，CARAFE)上采样算子替换YOLOv7中最邻近插值方式的上采样算子，从而提高了目标检测精度;引入分布移位卷积(distribution shifting convolution，DSConv)模块替换YOLOv7的头部网络中部分传统卷积，实现了模型的轻量化。结果表明，改进算法的mAP值达到了 90。7％，模型计算量仅为96G。该方法具有准确率高、稳健性强等特点，在建筑垃圾分类检测实际场景中具有较高的应用价值。
Construction waste classification and detection algorithm based on improved YOLOv7
The construction waste classification and detection technology based on deep learning algorithm is of great significance to the construction waste recycling and resource reuse.In this paper,the improved YOLOv7 algorithm was proposed to realize the classification and detection of construction waste.Firstly,the content aware reassembly of features(CARAFE)upsampling operator was used to replace the nearest interpolation upsampling operator in YOLOv7,which reduced the loss of image quality in the upsampling process and aggregates contextual information within a large receptive field,thereby improving the detection accuracy of construction waste.Secondly,the distribution shifting convolution(DSConv)module was introduced to replace the traditional convolution in the head network of YOLOv7,achieving lightweight of the model.The experimental results showed that the mAP value of the improved model reached 90.7％,and the computational complexity was only 96G.The improved model had higher accuracy and stronger robustness performance.It has high application value in the field of construction waste classification and detection.
国家科技期刊平台
NSTL
万方数据
