class MainWindow(tk.Frame):
def __init__(self, master):
super(MainWindow, self).__init__(master)
self.pack()
self.master.title("Character Recognizer")
self.aspect_ratio = self.calculate_aspect_ratio()
self.drawing_canvas = tk.Canvas(self,
width=320,
height=180,
bg='black')
self.drawing_canvas.pack()
self.drawing_canvas.bind('<B1-Motion>', self.move)
self.bottom_frame = tk.Frame(self)
self.bottom_frame.pack(side='bottom', padx=15, pady=15)
self.button = tk.Button(self.bottom_frame, text="Process")
self.button.pack(side='left', padx=15, pady=15)
self.button.bind('<Button-1>', self.make_prediction)
self.clear_button = tk.Button(self.bottom_frame, text="Clear")
self.clear_button.pack(side='left', padx=15, pady=15)
self.clear_button.bind('<Button-1>', self.clear_canvas)
self.result = tk.Label(self.bottom_frame, text="Result")
self.result.pack(side='left', padx=15, pady=15)
self.prev = None
self.image_id = 0
self.get_predictor()
def calculate_aspect_ratio(self):
screen_width = self.master.winfo_screenwidth()
screen_height = self.master.winfo_screenheight()
aspect_ratio = (screen_width * 1.0) / screen_height
print("Width: {}, Height: {}, Aspect Ratio: {}".format(
screen_width, screen_height, aspect_ratio))
return aspect_ratio
def get_predictor(self):
self.predictor = Predictor()
def clear_canvas(self, event):
self.drawing_canvas.delete('all')
def extract_digit(self, img):
row_vals, col_vals = np.where(img == 255)
min_row = np.min(row_vals)
max_row = np.max(row_vals)
min_col = np.min(col_vals)
max_col = np.max(col_vals)
cropped_image = img[min_row - 10:max_row + 10,
min_col - 20:max_col + 20]
return cropped_image
def make_prediction(self, event):
self.image_id += 1
image_path = os.path.join(IMAGE_FOLDER,
'img' + str(self.image_id) + '.png')
x = self.master.winfo_rootx() + self.drawing_canvas.winfo_x()
y = self.master.winfo_rooty() + self.drawing_canvas.winfo_y()
x1 = x + self.drawing_canvas.winfo_width()
y1 = y + self.drawing_canvas.winfo_height()
img = ImageGrab.grab().crop((x, y, x1, y1)).convert("RGB")
image = np.array(img)
image = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
image = self.extract_digit(image)
#image = cv2.resize(image, (image.shape[1], int(image.shape[1] / self.aspect_ratio)))
cv2.imwrite(image_path, image)
print("Numpy array shape is: {}".format(image.shape))
prediction = self.predictor.predict(image)
self.result.config(text=str(prediction))
def move(self, event):
buffer_width = 5
x1, y1 = event.x - buffer_width, event.y - buffer_width
x2, y2 = event.x + buffer_width, event.y + buffer_width
self.drawing_canvas.create_oval(x1,
y1,
x2,
y2,
outline='white',
fill='white')
def perform_cleanup(self):
self.predictor.close_session()
self.delete_temporary_files()
def delete_temporary_files(self):
for image in os.listdir(IMAGE_FOLDER):
image_path = os.path.join(IMAGE_FOLDER, image)
if os.path.isfile(image_path):
os.unlink(image_path)
