【目标检测】利用PyQT5搭建YOLOv5可视化界面

应广大读者需求，重构了整个仓库，目前适配YOLOv5最新版本。

开源地址：https://github.com/zstar1003/yolov5_pyqt5

最新界面：

目前支持图像/视频/摄像头检测，适配YOLOv5各版本模型。

---

本篇主要利用PyQT5搭建YOLOv5可视化界面，并打包成exe程序。

整体框架参考自：https://xugaoxiang.com/2021/06/30/yolov5-pyqt5
在此基础上，优化了预测逻辑，适配YOLOv5-5.0版本，并使用qdarkstyle美化了界面，支持图片检测、摄像头检测、视频检测，整体效果如下图所示：

开源仓库：https://github.com/zstar1003/yolov5_pyqt5
可直接运行的exe程序：https://pan.baidu.com/s/16nHvS5tRSeLKB0Ql2-6ZFw?pwd=8888

项目整体框架如下图所示：

· models：存放模型构建相关程序，直接从yolov5-5.0版本中clone过来

• utils：存放绘图、数据加载等相关工具，直接从yolov5-5.0版本中clone过来
• UI：存放软件图标
• result：存放预测之后的图片或视频
• weights：模型权重，默认使用YOLOv5官方提供的yolov5s.pt

main.py

import osimport sysimport cv2import randomimport torchimport numpy as npimport torch.backends.cudnn as cudnnimport qdarkstylefrom PyQt5 import QtCore, QtGui, QtWidgetsfrom PyQt5.QtGui import QIcon, QPixmapfrom models.experimental import attempt_loadfrom utils.general import check_img_size, non_max_suppression, scale_coordsfrom utils.datasets import letterboxfrom utils.plots import plot_one_boxclass Ui_MainWindow(QtWidgets.QMainWindow):    def __init__(self, parent=None):        super(Ui_MainWindow, self).__init__(parent)        self.timer_video = QtCore.QTimer()        self.setupUi(self)        self.init_logo()        self.init_slots()        self.cap = cv2.VideoCapture()        self.out = None        self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")        self.half = self.device.type != 'cpu'  # half precision only supported on CUDA        cudnn.benchmark = True        weights = 'weights/yolov5s.pt'   # 模型加载路径        imgsz = 640  # 预测图尺寸大小        self.conf_thres = 0.25  # NMS置信度        self.iou_thres = 0.45  # IOU阈值        # 载入模型        self.model = attempt_load(weights, map_location=self.device)        stride = int(self.model.stride.max())        self.imgsz = check_img_size(imgsz, s=stride)        if self.half:            self.model.half()  # to FP16        # 从模型中获取各类别名称        self.names = self.model.module.names if hasattr(self.model, 'module') else self.model.names        # 给每一个类别初始化颜色        self.colors = [[random.randint(0, 255) for _ in range(3)] for _ in self.names]    def setupUi(self, MainWindow):        MainWindow.setObjectName("MainWindow")        MainWindow.resize(900, 600)        # MainWindow.setStyleSheet("")        self.centralwidget = QtWidgets.QWidget(MainWindow)        self.centralwidget.setObjectName("centralwidget")        # self.centralwidget.setStyleSheet("border: 1px solid white;")        self.horizontalLayout_2 = QtWidgets.QHBoxLayout(self.centralwidget)        self.horizontalLayout_2.setObjectName("horizontalLayout_2")        self.horizontalLayout = QtWidgets.QHBoxLayout()        self.horizontalLayout.setSizeConstraint(QtWidgets.QLayout.SetNoConstraint)        self.horizontalLayout.setObjectName("horizontalLayout")        self.verticalLayout = QtWidgets.QVBoxLayout()        self.verticalLayout.setContentsMargins(0, 0, 0, 0)  # 布局的左、上、右、下到窗体边缘的距离        # self.verticalLayout.setSpacing(0)        self.verticalLayout.setObjectName("verticalLayout")        # 打开图片按钮        self.pushButton_img = QtWidgets.QPushButton(self.centralwidget)        sizePolicy = QtWidgets.QSizePolicy(QtWidgets.QSizePolicy.Minimum, QtWidgets.QSizePolicy.MinimumExpanding)        sizePolicy.setHorizontalStretch(0)        sizePolicy.setVerticalStretch(0)        sizePolicy.setHeightForWidth(self.pushButton_img.sizePolicy().hasHeightForWidth())        self.pushButton_img.setSizePolicy(sizePolicy)        self.pushButton_img.setMinimumSize(QtCore.QSize(150, 40))        self.pushButton_img.setMaximumSize(QtCore.QSize(150, 40))        font = QtGui.QFont()        font.setFamily("Agency FB")        font.setPointSize(12)        self.pushButton_img.setFont(font)        self.pushButton_img.setObjectName("pushButton_img")        self.verticalLayout.addWidget(self.pushButton_img, 0, QtCore.Qt.AlignHCenter)        self.verticalLayout.addStretch(5)  # 增加垂直盒子内部对象间距        # 打开摄像头按钮        self.pushButton_camera = QtWidgets.QPushButton(self.centralwidget)        sizePolicy = QtWidgets.QSizePolicy(QtWidgets.QSizePolicy.Minimum, QtWidgets.QSizePolicy.Expanding)        sizePolicy.setHorizontalStretch(0)        sizePolicy.setVerticalStretch(0)        sizePolicy.setHeightForWidth(self.pushButton_camera.sizePolicy().hasHeightForWidth())        self.pushButton_camera.setSizePolicy(sizePolicy)        self.pushButton_camera.setMinimumSize(QtCore.QSize(150, 40))        self.pushButton_camera.setMaximumSize(QtCore.QSize(150, 40))        self.pushButton_camera.setFont(font)        self.pushButton_camera.setObjectName("pushButton_camera")        self.verticalLayout.addWidget(self.pushButton_camera, 0, QtCore.Qt.AlignHCenter)        self.verticalLayout.addStretch(5)        # 打开视频按钮        self.pushButton_video = QtWidgets.QPushButton(self.centralwidget)        sizePolicy = QtWidgets.QSizePolicy(QtWidgets.QSizePolicy.Minimum, QtWidgets.QSizePolicy.Expanding)        sizePolicy.setHorizontalStretch(0)        sizePolicy.setVerticalStretch(0)        sizePolicy.setHeightForWidth(self.pushButton_video.sizePolicy().hasHeightForWidth())        self.pushButton_video.setSizePolicy(sizePolicy)        self.pushButton_video.setMinimumSize(QtCore.QSize(150, 40))        self.pushButton_video.setMaximumSize(QtCore.QSize(150, 40))        self.pushButton_video.setFont(font)        self.pushButton_video.setObjectName("pushButton_video")        self.verticalLayout.addWidget(self.pushButton_video, 0, QtCore.Qt.AlignHCenter)        self.verticalLayout.addStretch(50)        # 显示导出文件夹按钮        self.pushButton_showdir = QtWidgets.QPushButton(self.centralwidget)        sizePolicy = QtWidgets.QSizePolicy(QtWidgets.QSizePolicy.Minimum, QtWidgets.QSizePolicy.Expanding)        sizePolicy.setHorizontalStretch(0)        sizePolicy.setVerticalStretch(0)        sizePolicy.setHeightForWidth(self.pushButton_showdir.sizePolicy().hasHeightForWidth())        self.pushButton_showdir.setSizePolicy(sizePolicy)        self.pushButton_showdir.setMinimumSize(QtCore.QSize(150, 50))        self.pushButton_showdir.setMaximumSize(QtCore.QSize(150, 50))        self.pushButton_showdir.setFont(font)        self.pushButton_showdir.setObjectName("pushButton_showdir")        self.verticalLayout.addWidget(self.pushButton_showdir, 0, QtCore.Qt.AlignHCenter)        # 右侧图片/视频填充区域        self.verticalLayout.setStretch(2, 1)        self.horizontalLayout.addLayout(self.verticalLayout)        self.label = QtWidgets.QLabel(self.centralwidget)        self.label.setObjectName("label")        self.horizontalLayout.addWidget(self.label)        self.horizontalLayout.setStretch(0, 1)        self.horizontalLayout.setStretch(1, 3)        self.horizontalLayout_2.addLayout(self.horizontalLayout)        self.label.setStyleSheet("border: 1px solid white;")  #  添加显示区域边框        # 底部美化导航条        MainWindow.setCentralWidget(self.centralwidget)        self.menubar = QtWidgets.QMenuBar(MainWindow)        self.menubar.setGeometry(QtCore.QRect(0, 0, 800, 23))        self.menubar.setObjectName("menubar")        MainWindow.setMenuBar(self.menubar)        self.statusbar = QtWidgets.QStatusBar(MainWindow)        self.statusbar.setObjectName("statusbar")        MainWindow.setStatusBar(self.statusbar)        self.retranslateUi(MainWindow)        QtCore.QMetaObject.connectSlotsByName(MainWindow)    def retranslateUi(self, MainWindow):        _translate = QtCore.QCoreApplication.translate        MainWindow.setWindowTitle(_translate("MainWindow", "YOLOv5目标检测平台"))        self.pushButton_img.setText(_translate("MainWindow", "图片检测"))        self.pushButton_camera.setText(_translate("MainWindow", "摄像头检测"))        self.pushButton_video.setText(_translate("MainWindow", "视频检测"))        self.pushButton_showdir.setText(_translate("MainWindow", "打开输出文件夹"))        self.label.setText(_translate("MainWindow", "TextLabel"))    def init_slots(self):        self.pushButton_img.clicked.connect(self.button_image_open)        self.pushButton_video.clicked.connect(self.button_video_open)        self.pushButton_camera.clicked.connect(self.button_camera_open)        self.pushButton_showdir.clicked.connect(self.button_show_dir)        self.timer_video.timeout.connect(self.show_video_frame)    def init_logo(self):        pix = QtGui.QPixmap('')   # 绘制初始化图片        self.label.setScaledContents(True)        self.label.setPixmap(pix)    def button_image_open(self):        print('打开图片')        name_list = []        img_name, _ = QtWidgets.QFileDialog.getOpenFileName(            self, "打开图片", "", "*.jpg;;*.png;;All Files(*)")        if not img_name:            return        img = cv2.imread(img_name)        print(img_name)        showimg = img        with torch.no_grad():            img = letterbox(img, new_shape=self.imgsz)[0]            # Convert            # BGR to RGB, to 3x416x416            img = img[:, :, ::-1].transpose(2, 0, 1)            img = np.ascontiguousarray(img)            img = torch.from_numpy(img).to(self.device)            img = img.half() if self.half else img.float()  # uint8 to fp16/32            img /= 255.0  # 0 - 255 to 0.0 - 1.0            if img.ndimension() == 3:                img = img.unsqueeze(0)            # Inference            pred = self.model(img)[0]            # Apply NMS            pred = non_max_suppression(pred, self.conf_thres, self.iou_thres)            # Process detections            for i, det in enumerate(pred):                if det is not None and len(det):                    # Rescale boxes from img_size to im0 size                    det[:, :4] = scale_coords(                        img.shape[2:], det[:, :4], showimg.shape).round()                    for *xyxy, conf, cls in reversed(det):                        label = '%s %.2f' % (self.names[int(cls)], conf)                        # print(label.split()[0])  # 打印各目标名称                        name_list.append(self.names[int(cls)])                        plot_one_box(xyxy, showimg, label=label,         color=self.colors[int(cls)], line_thickness=2)        cv2.imwrite('result/prediction.jpg', showimg)        self.result = cv2.cvtColor(showimg, cv2.COLOR_BGR2BGRA)        self.result = cv2.resize(self.result, (640, 480), interpolation=cv2.INTER_AREA)        self.QtImg = QtGui.QImage(self.result.data, self.result.shape[1], self.result.shape[0], QtGui.QImage.Format_RGB32)        self.label.setPixmap(QtGui.QPixmap.fromImage(self.QtImg))    def button_video_open(self):        video_name, _ = QtWidgets.QFileDialog.getOpenFileName(            self, "打开视频", "", "*.mp4;;*.avi;;All Files(*)")        if not video_name:            return        flag = self.cap.open(video_name)        if flag == False:            QtWidgets.QMessageBox.warning(                self, u"Warning", u"打开视频失败", buttons=QtWidgets.QMessageBox.Ok, defaultButton=QtWidgets.QMessageBox.Ok)        else:            self.out = cv2.VideoWriter('result/vedio_prediction.avi', cv2.VideoWriter_fourcc(                *'MJPG'), 20, (int(self.cap.get(3)), int(self.cap.get(4))))            self.timer_video.start(30)            self.pushButton_video.setDisabled(True)            self.pushButton_img.setDisabled(True)            self.pushButton_camera.setDisabled(True)    def button_camera_open(self):        if not self.timer_video.isActive():            # 默认使用第一个本地camera            flag = self.cap.open(0)            if flag == False:                QtWidgets.QMessageBox.warning(                    self, u"Warning", u"打开摄像头失败", buttons=QtWidgets.QMessageBox.Ok, defaultButton=QtWidgets.QMessageBox.Ok)            else:                self.out = cv2.VideoWriter('result/camera_prediction.avi', cv2.VideoWriter_fourcc(                    *'MJPG'), 20, (int(self.cap.get(3)), int(self.cap.get(4))))                self.timer_video.start(30)                self.pushButton_video.setDisabled(True)                self.pushButton_img.setDisabled(True)                self.pushButton_camera.setText(u"关闭摄像头")        else:            self.timer_video.stop()            self.cap.release()            self.out.release()            self.label.clear()            self.init_logo()            self.pushButton_video.setDisabled(False)            self.pushButton_img.setDisabled(False)            self.pushButton_camera.setText(u"摄像头检测")    def show_video_frame(self):        name_list = []        flag, img = self.cap.read()        if img is not None:            showimg = img            with torch.no_grad():                img = letterbox(img, new_shape=self.imgsz)[0]                # Convert                # BGR to RGB, to 3x416x416                img = img[:, :, ::-1].transpose(2, 0, 1)                img = np.ascontiguousarray(img)                img = torch.from_numpy(img).to(self.device)                img = img.half() if self.half else img.float()  # uint8 to fp16/32                img /= 255.0  # 0 - 255 to 0.0 - 1.0                if img.ndimension() == 3:                    img = img.unsqueeze(0)                # Inference                pred = self.model(img)[0]                # Apply NMS                pred = non_max_suppression(pred, self.conf_thres, self.iou_thres)                # Process detections                for i, det in enumerate(pred):  # detections per image                    if det is not None and len(det):                        # Rescale boxes from img_size to im0 size                        det[:, :4] = scale_coords(img.shape[2:], det[:, :4], showimg.shape).round()                        # Write results                        for *xyxy, conf, cls in reversed(det):label = '%s %.2f' % (self.names[int(cls)], conf)name_list.append(self.names[int(cls)])# print(label)  # 打印各目标+置信度plot_one_box(    xyxy, showimg, label=label, color=self.colors[int(cls)], line_thickness=2)            self.out.write(showimg)            show = cv2.resize(showimg, (640, 480))            self.result = cv2.cvtColor(show, cv2.COLOR_BGR2RGB)            showImage = QtGui.QImage(self.result.data, self.result.shape[1], self.result.shape[0],         QtGui.QImage.Format_RGB888)            self.label.setPixmap(QtGui.QPixmap.fromImage(showImage))        else:            self.timer_video.stop()            self.cap.release()            self.out.release()            self.label.clear()            self.pushButton_video.setDisabled(False)            self.pushButton_img.setDisabled(False)            self.pushButton_camera.setDisabled(False)            self.init_logo()    def button_show_dir(self):        path = os.getcwd() + '\\' + 'result'        os.system(f"start explorer {path}")if __name__ == '__main__':    app = QtWidgets.QApplication(sys.argv)    app.setStyleSheet(qdarkstyle.load_stylesheet_pyqt5())    ui = Ui_MainWindow()    # 设置窗口透明度    # ui.setWindowOpacity(0.93)    # 去除顶部边框    # ui.setWindowFlags(Qt.FramelessWindowHint)    # 设置窗口图标    icon = QIcon()    icon.addPixmap(QPixmap("./UI/icon.ico"), QIcon.Normal, QIcon.Off)    ui.setWindowIcon(icon)    ui.show()    sys.exit(app.exec_())

整体逻辑是软件已启动就开始载入模型，然后利用槽函数去响应按钮信息。

为了尽可能减少打包之后的体积，在打包之前，先使用Anaconda新建一个虚拟环境并安装好pytorch等YOLOv5所需必要库。

打包通常采用的是Pyinstaller这个工具库，本次打包使用一个新的工具叫Auto Py to Exe，该工具仍是调用Pyinstaller进行打包，不过对选项进行了可视化，操作更加便捷。

安装方式：

git clone https://github.com/brentvollebregt/auto-py-to-exe.gitpython setup.py install 

注意安装时可能会提示缺少一些包，依次pip安装即可，geventwebsocket库需要这样进行安装。

pip install gevent-websocket

安装好之后，在终端输入auto-py-to-exe，会在浏览器中默认打开如下界面：

脚本位置选择main.py，选择单目录模式，隐藏控制台，并选择图标和输出路径，然后就可以一键进行打包。

打包完成之后，会在输出文件夹下输入一个main文件夹。
运行之前，需要将原始工程中的几个文件夹拷贝进去，否则会提示找不到文件，如下图所示：

双击main.exe，即可看到可视化界面。

在调式时，遇到一些小问题，这里也记录下。

问题一：遇到警告：

UserWarning: torch.meshgrid: in an upcoming release, it will be required to …

在报错的文件中将

return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]

修改为

return _VF.meshgrid(tensors, **kwargs, indexing = ‘ij’) # type: ignore[attr-defined]

问题二：
打包时遇到的错误：

ImportError: ERROR: recursion is detected during loading of “cv2” binary extensions. Check OpenCV installation.

pyinstaller和cv2版本存在兼容问题，卸载已有的opencv-python，安装opencv-python=4.5.3.56

来源地址：https://blog.csdn.net/qq1198768105/article/details/126463741