Flashing console when running exe packaged by pyinstaller

[WISHF-Amoy]

I use YOLO8 to make a detection program, and use pyqt5 make the GUI. I packaged program using pyinstaller, when I run the exe without starting YOLO8 detection, evrtthing goes right, but in the first time I click the button to start detection task, the console which I packged with option -w will come out and strangly flashing twice, and then wait for a longer time than run in PyCharm, the detection begin. So I just want to know how to avoid this strange flash.

Below is the pseudocode of my program:

import multiprocessing

def GetPredict(screenshot):
    results = model(screenshot)
    for i, r in enumerate(results):
        # Plot results image
        im_bgr = r.plot()  # BGR-order numpy array
        im_rgb = Image.fromarray(im_bgr[..., ::-1])  # RGB-order PIL image
        
        return xxx


class Window(QMainWindow):
    def __init__(self):
        super().__init__()
        
        self.ui = Ui_MainWindow()
        self.ui.setupUi(self)
        self.setWindowFlags(Qt.FramelessWindowHint)
        self.setAttribute(Qt.WA_TranslucentBackground)

        self.timer = QTimer()
        self.timer.timeout.connect(self.LoopFunction)

        self.ui.Start_Button.clicked.connect(self.start_loop)
        self.ui.Stop_Button.clicked.connect(self.stop_loop)

    def LoopFunction(self):
        screnshot = xxx
        result = GetPredict(screenshot)

    def start_loop(self):
        self.timer.start(1000)  # tick

    def stop_loop(self):
        self.timer.stop()

if __name__ == '__main__':
    # added
    multiprocessing.freeze_support()

    app = QApplication(sys.argv)
    window = Window()
    window.show()
    sys.exit(app.exec_())

[rokm]

Sounds like your code (or the 3rd party code you are using) is spawning subprocesses with programs/command that require console. And since the parent application does not have a console, each such subprocesses opens their own console. If you are using ultralytics yolov8 on CPU, then this device auto-selection codepath might be the culprit:

https://github.com/ultralytics/ultralytics/blob/911a0edd2d1b47c28e1eef657859dbb9bcffb13f/ultralytics/utils/torch_utils.py#L164
https://github.com/ultralytics/ultralytics/blob/911a0edd2d1b47c28e1eef657859dbb9bcffb13f/ultralytics/utils/torch_utils.py#L73

It ends up using cpuinfo.get_cpu_info, which seems to do several subprocess calls, e.g. to wmi on Windows.

Nothing we can do about that; maybe explicitly specifying the device to suppress the auto-selection will help.

---

Note that if you call multiprocessing.freeze_support() like that, you are probably unnecessarily importing PyQt5 packages in multiprocessing helpers, slowing them down.

See ##8211 and https://pyinstaller.org/en/stable/common-issues-and-pitfalls.html#when-to-call-multiprocessing-freeze-support

[WISHF-Amoy]

Oh! Your advice explicitly specifying the device really works! Thanks a lot for your help!