工业焊接焊缝缺陷检测数据集-焊接缺陷检测数据集8876张存在数据增强处理提供yolo和voc两种标注方式10类标注数量Overlap: 173 — 重叠Spatter: 2491 — 飞溅Welding_line: 10738 — 焊接线Porosity: 4023 — 孔隙Undercut: 319 — 缺口Crack: 1434 — 裂纹Overfill: 494 — 过填Weld Bead Irregularity: 433 — 焊缝不规则Burn-through: 449 — 穿透Crater: 449 — 坑洞Image num: 8876焊接缺陷检测数据集数据集总览表项目参数数据集名称焊接缺陷检测数据集图像总数量8876张含数据增强标注格式YOLO、VOC类别数量10类类别明细序号英文标签中文名称标注实例数0Overlap重叠1731Spatter飞溅24912Welding_line焊接线107383Porosity孔隙40234Undercut缺口3195Crack裂纹14346Overfill过填4947Weld Bead Irregularity焊缝不规则4338Burn‑through穿透4499Crater坑洞449yolo配置文件 weld.yamltrain:./train/imagesval:./val/imagestest:./test/imagesnc:10names:0:Overlap1:Spatter2:Welding_line3:Porosity4:Undercut5:Crack6:Overfill7:Weld Bead Irregularity8:Burn‑through9:Crater关键词焊接缺陷检测、焊缝质检、工业视觉检测、机器视觉、钢结构焊接检测、工业缺陷识别、YOLO目标检测、焊接裂纹孔隙识别标签#焊接缺陷数据集#工业质检#焊缝检测#机器视觉数据集应用场景钢结构工厂焊缝自动化质检流水线自动识别焊接飞溅、裂纹、孔隙、坑洞等缺陷替代人工肉眼复检提升生产效率。压力容器、管道焊接质量检测识别穿透、缺口、过填等高危焊接缺陷规避设备安全隐患。轨道交通、工程机械零部件质检焊接工件视觉筛查保障构件焊接强度。科研算法实验工业小缺陷检测模型训练、消融对比实验适合YOLO系列改进算法验证。焊接机器人配套视觉系统实时采集焊缝图像在线识别焊接瑕疵辅助机器人调整焊接工艺。YOLOv11训练代码 train_weld.pyfromultralyticsimportYOLOif__name____main__:modelYOLO(yolo11n.pt)resultsmodel.train(dataweld.yaml,epochs60,imgsz640,batch12,device0,workers2,conf0.25)model.val()#单图推理model.predict(source./test.jpg,saveTrue)推理代码 infer_weld.pyfromultralyticsimportYOLO modelYOLO(best.pt)defweld_detect(img_path):resmodel.predict(img_path,conf0.25)forrinres:forboxinr.boxes:cls_namemodel.names[int(box.cls)]conffloat(box.conf)print(f缺陷类别:{cls_name}置信度:{conf:.2f})res[0].save(weld_result.jpg)if__name____main__:weld_detect(test.jpg)PyQt5简易可视化界面 weld_gui.pyimportsysfromPyQt5.QtWidgetsimport(QApplication,QMainWindow,QWidget,QVBoxLayout,QPushButton,QLineEdit,QTableWidget,QTableWidgetItem,QFileDialog,QLabel)fromPyQt5.QtGuiimportQPixmapfromultralyticsimportYOLOclassWeldDetectWindow(QMainWindow):def__init__(self):super().__init__()self.setWindowTitle(焊接缺陷检测系统)self.resize(1050,720)self.modelYOLO(best.pt)self.img_pathNoneself.init_ui()definit_ui(self):centralQWidget()self.setCentralWidget(central)layoutQVBoxLayout(central)self.img_labelQLabel(焊缝图像显示)layout.addWidget(self.img_label)self.line_editQLineEdit()layout.addWidget(self.line_edit)self.btn_openQPushButton(打开焊缝图片)self.btn_open.clicked.connect(self.open_image)layout.addWidget(self.btn_open)self.btn_detectQPushButton(执行缺陷检测)self.btn_detect.clicked.connect(self.run_detect)layout.addWidget(self.btn_detect)self.tableQTableWidget()self.table.setColumnCount(3)self.table.setHorizontalHeaderLabels([序号,缺陷类型,置信度])layout.addWidget(self.table)defopen_image(self):f,_QFileDialog.getOpenFileName(filter图片(*.jpg *.png))iff:self.img_pathf self.line_edit.setText(f)self.img_label.setPixmap(QPixmap(f).scaled(self.img_label.size()))defrun_detect(self):ifnotself.img_path:returnresself.model.predict(self.img_path,conf0.25)self.table.setRowCount(0)index0forrinres:forboxinr.boxes:index1clsself.model.names[int(box.cls)]confround(float(box.conf),2)rowself.table.rowCount()self.table.insertRow(row)self.table.setItem(row,0,QTableWidgetItem(str(index)))self.table.setItem(row,1,QTableWidgetItem(cls))self.table.setItem(row,2,QTableWidgetItem(str(conf)))r.save(weld_out.jpg)if__name____main__:appQApplication(sys.argv)winWeldDetectWindow()win.show()sys.exit(app.exec_())环境依赖pip install ultralytics torch opencv-python pyqt5将训练输出best.pt放到代码同级目录直接运行weld_gui.py启动软件。