智慧化-桥梁表面五类病害检测与分割数据集数据集面向桥梁结构表观病害识别、像素级分割与智能巡检任务覆盖裂缝、腐蚀、析白、渗漏水和混凝土剥落五种常见损伤类型。数据同时提供原始图像、像素标签、标签信息和样本清单可用于语义分割、病害区域提取、损伤面积评估及巡检算法研究。图像数量5,821张腐蚀图像955张裂缝图像1,480张析白图像1,269张渗漏水图像873张剥落图像1,244张类别数量5类病害另含背景标签标注类型像素级语义分割标签配套文件原始图像、标签图、JSON标签说明、样本列表数据体积约2.07G适用任务桥梁病害分割、结构巡检、损伤区域定位、工程视觉分析梁表面五类病害检测与分割数据集数据集信息表项目详细说明数据集名称梁表面五类病害检测与分割数据集图像总量5821张各类病害样本分布腐蚀955张裂缝1480张析白1269张渗漏水873张混凝土剥落1244张分割类别5类病害 背景crack裂缝、corrosion腐蚀、efflorescence析白、water_seepage渗漏水、spalling混凝土剥落、background背景标注类型像素级语义分割标注配套文件原始图像、分割标签图、JSON标签说明、样本清单数据大小约2.07GB适配任务桥梁表观病害语义分割、病害区域提取、损伤面积测算、桥梁结构智能巡检、工程结构视觉算法研发推荐模型YOLOv8-seg / YOLOv11-seg、U-Net、SegFormer、Mask R-CNN分割任务yaml配置YOLO-seg 标准格式 bridge_defect_seg.yamlpath:./bridge_surface_defect_datasettrain:images/trainval:images/valtest:images/testnc:5names:0:crack1:corrosion2:efflorescence3:water_seepage4:spalling环境依赖pipinstallultralytics torch opencv-python pillow numpyYOLOv8-seg 训练代码 train_bridge_seg.pyfromultralyticsimportYOLOdefmain():# 加载分割预训练权重兼容 YOLOv8-seg / YOLOv11-segmodelYOLO(yolov8s-seg.pt)train_resultmodel.train(data./bridge_surface_defect_dataset/bridge_defect_seg.yaml,epochs100,imgsz640,batch12,device0,workers4,patience15,projectruns/train,namebridge_defect_segment,exist_okTrue)print(训练完成权重保存路径runs/train/bridge_defect_segment/weights)metricsmodel.val(splittest)print(fmAP50:{metrics.box.map50:.4f})if__name____main__:main()分割推理代码 predict_bridge_seg.pyfromultralyticsimportYOLOimportcv2 modelYOLO(runs/train/bridge_defect_segment/weights/best.pt)if__name____main__:img_filebridge_test.jpgresultsmodel.predict(sourceimg_file,conf0.25,saveTrue,save_masksTrue# 输出单独病害掩码图用于面积计算)vis_imgresults[0].plot()cv2.imshow(Bridge Defect Segmentation,vis_img)cv2.waitKey(0)cv2.destroyAllWindows()关键词标签#桥梁病害分割#混凝土裂缝检测#桥梁智能巡检#语义分割数据集#结构表观损伤识别#YOLO-seg数据集中英术语对照表标注通用中文病害名称标准英文标签裂缝crack腐蚀corrosion析白泛碱efflorescence渗漏水water_seepage混凝土剥落spalling