Matplotlib affecting tkinter window size and process behavior

Question

I'm trying to build a GUI with tkinter that displays an image which is dynamically updated based on the inputs given (which i removed from the MRE below).

When I generate a dummy image like np.random.randint(255, size=(2000, 2000, 3), dtype=np.uint8), the GUI works as expected and when i close it with the classic close button (red cross), the process terminates as expected. However, when I use the actual function to produce the type of image I want (generate_voronoi, which loads an input image from its file name and edits it via matplotlib), the whole GUI appears smaller and when I close the window, the process in the shell appears to be still running (no prompt). This is the code for the GUI:

class Example(Frame):

    def __init__(self, master):
        super().__init__(master)
        self._init_image_gui()
        self._set_default_params()
        self._update_img()

    def _init_image_gui(self):
        # Create an object of tkinter ImageTk
        self.img = Image.open("jap.jpg").resize((300,300))
        self.img = ImageTk.PhotoImage(self.img)
        # Create a Label Widget to display the text or Image
        self.img_label = tk.Label(image=self.img)
        self.img_label.grid(column=0, row=3, columnspan=4)

    def _update_img(self):
        # produce image
        #voronoi_img = np.random.randint(255, size=(2000, 2000, 3), dtype=np.uint8)
        voronoi_img = generate_voronoi(self.input_path, self.output_path, self.n_regions, self.padding_amount, self.pad_color, self.rounding_amount)
        voronoi_img = voronoi_img.astype(np.uint8)
        voronoi_img = cv2.cvtColor(voronoi_img, cv2.COLOR_BGR2RGB)
        new_img = Image.fromarray(voronoi_img).resize((200, 200))
        new_img = ImageTk.PhotoImage(new_img)
        # update the UI to show it
        self.img_label.configure(image=new_img)
        self.img = new_img

    def _set_default_params(self):
        self.input_path = "jap.jpg"
        self.output_path = "jap_out.jpg"
        self.n_regions = 100
        self.padding_amount = 10
        self.pad_color = "#000000"
        self.rounding_amount = 10


def main():
    root = Tk()
    root.geometry("500x500")
    app = Example(root)
    root.mainloop()

Altough I think the problem is in setting up the ax to work with matplotlib without any frames or paddings in the output image (which I then convert to numpy array), that is called by generate_voronoi, and is this:

def setup_plot(x : int, y : int, pad_color : str):
    '''Sets up matplotlib paint axes and returns it'''

    _, ax = plt.subplots(figsize=(x / 100, y / 100))
    ax.axis([0, x, 0, y])
    ax.axis('off')
    ax.set_facecolor(pad_color)
    ax.add_artist(ax.patch)
    ax.patch.set_zorder(-1)

    # setup ax size in pixels
    ax.set_position([0, 0, 1, 1])
    return ax

How can I achieve the same correct behavior as with the "dummy" image generation?

TLDR

Why does my GUI shrink in size and the process doesn't terminate when I add just this matplotlib commands (without even using them)?

def _update_img(self):
    plt.figure() # this causes the problem
    voronoi_img = np.random.randint(255, size=(2000, 2000, 3)

Full code can be found here

Answer 1

I solved it this way:

target_df['FiscalCode'] = target_df.apply(lambda x: codicefiscale.encode(x["Surname"], x["Name"], x["Sex"], x["BirthDate"], x["BirthPlace"]), axis=1)