def test_formula_expressions_can_be_parsed():
formula = parser.expr(
'base_value * MULTIPLIER + STAMINA // 3 - SPEED // 5 + ENHANCERS'
).compile()
value = eval(formula, {}, {
'base_value': 3,
'MULTIPLIER': 1,
'STAMINA': 3,
'SPEED': 10,
'ENHANCERS': 3
})
assert value == 5
# expr = NumExpr('base_value * MULTIPLIER + STAMINA / 3 - SPEED / 5 + ENHANCERS')
variables = {
'base_value': 3,
'MULTIPLIER': 1,
'STAMINA': 3,
'SPEED': 10,
'ENHANCERS': 3,
'BANANAS': 8
}
formula = Formula(
'base_value * MULTIPLIER + STAMINA // 3 - SPEED // 5 + ENHANCERS')
assert formula.evaluate(variables) == 5
class MathReader(QWidget):
def __init__(self):
super().__init__()
self.iterations = 10000
self.ro = 0.5
self.alpha = 0.5
self.fonts = Fonts(3)
fonts_used = self.fonts.fonts
self.fonts.load()
self.network = Network(neurons=(35, 25, len(Constants.symbols)),
input_size=21 * 14)
self.formula = Formula()
self.file_path = ""
self.network_ready = False
self.setWindowTitle("MathReader")
self.setWindowIcon(QIcon('res/icon.png'))
grid = QGridLayout()
self.setFixedSize(730, 500)
label_math_reader = QLabel("MathReader")
label_math_reader.setFont(QFont("Ink Free", 20, QFont.Bold))
label_math_reader.setMaximumHeight(50)
label_math_reader.setAlignment(Qt.AlignCenter)
self.btn_save_weights = QPushButton("Save weights")
self.btn_save_weights.clicked.connect(self.save_weights)
self.btn_load_weights = QPushButton("Load weights")
self.btn_load_weights.clicked.connect(self.load_weights)
self.btn_train = QPushButton("Train")
self.btn_train.clicked.connect(self.open_training_dialog)
self.btn_browse = QPushButton("Browse")
self.btn_browse.clicked.connect(self.open_file)
self.btn_solve = QPushButton("Solve!")
self.btn_solve.setEnabled(False)
self.btn_solve.clicked.connect(self.solve_formula)
label_file = QLabel("Chosen image directory: ")
label_open_image = QLabel(
"Use the 'Browse' button to select graphic file, you would like me to read from :)"
)
label_open_image.setAlignment(Qt.AlignCenter)
self.label_image = QLabel("Your graphic file will load here!")
self.label_image.setStyleSheet("background-color: white")
self.label_image.setAlignment(Qt.AlignCenter)
self.browsed_file = QLineEdit()
self.browsed_file.setReadOnly(True)
self.browsed_file.setMaximumWidth(300)
self.solved_formula = QLineEdit()
self.solved_formula.setReadOnly(True)
self.solved_formula.setFont(QFont("Bahnschrift", 12))
self.solved_formula.setAlignment(Qt.AlignCenter)
label_solved = QLabel("I think it is...")
label_solved.setMaximumHeight(20)
label_solved.setAlignment(Qt.AlignBottom)
self.loading = QProgressBar()
label_loading = QLabel("Training state: ")
self.separator_h = QFrame()
self.separator_h.setFrameShape(QFrame.HLine)
self.separator_h.setFrameShadow(QFrame.Sunken)
separator_v = QFrame()
separator_v.setFrameShape(QFrame.VLine)
separator_v.setFrameShadow(QFrame.Sunken)
start_state = "Loaded " + str(fonts_used) + " fonts"
start_state += ("\nInitializing network with layers of sizes " +
str(self.network.neurons))
self.label_state = QLabel(start_state)
self.label_state.setStyleSheet("background-color: white")
self.label_state.setAlignment(Qt.AlignCenter)
self.training_dialog = QDialog()
self.training_params = QLabel(
"Here you can specify training parameters: ")
self.choose_iterations_label = QLabel("Iterations number: ")
self.choose_alpha_label = QLabel("\u03B1 value: ")
self.choose_rho_label = QLabel("\u03C1 value: ")
self.choose_iterations_input = QLineEdit()
self.choose_iterations_input.setText("10000")
self.choose_alpha_input = QLineEdit()
self.choose_alpha_input.setText("0.5")
self.choose_rho_input = QLineEdit()
self.choose_rho_input.setText("0.5")
self.btn_apply = QPushButton("Apply")
self.btn_apply.setMaximumWidth(80)
grid.addWidget(label_math_reader, 0, 0, 1, 1)
grid.addWidget(self.btn_train, 1, 0)
grid.addWidget(self.btn_save_weights, 2, 0)
grid.addWidget(self.btn_load_weights, 3, 0)
grid.addWidget(self.btn_solve, 5, 0)
grid.addWidget(separator_v, 0, 1, 10, 1)
grid.addWidget(label_open_image, 0, 2, 1, 3)
grid.addWidget(label_file, 1, 2, 1, 1)
grid.addWidget(self.browsed_file, 1, 3, 1, 1)
grid.addWidget(self.btn_browse, 1, 4, 1, 1)
grid.addWidget(self.label_image, 2, 2, 7, 3)
grid.addWidget(label_loading, 9, 2, Qt.AlignCenter)
grid.addWidget(self.loading, 9, 3, 1, 2)
grid.addWidget(label_solved, 10, 0, 1, 1)
grid.addWidget(self.solved_formula, 11, 0, 1, 5)
grid.addWidget(self.label_state, 12, 0, 1, 5)
grid.setSpacing(10)
self.setLayout(grid)
self.training_dialog.setModal(True)
self.training_dialog.setWindowTitle("Training Dialog")
self.training_dialog.setWindowIcon(QIcon('res/icon.png'))
grid = QGridLayout()
self.btn_apply.clicked.connect(self.get_training_params)
self.training_dialog.setFixedSize(230, 150)
grid.addWidget(self.training_params, 0, 0, 1, 2)
grid.addWidget(self.separator_h, 1, 0, 1, 2)
grid.addWidget(self.choose_iterations_label, 2, 0)
grid.addWidget(self.choose_iterations_input, 2, 1)
grid.addWidget(self.choose_alpha_label, 3, 0)
grid.addWidget(self.choose_alpha_input, 3, 1)
grid.addWidget(self.choose_rho_label, 4, 0)
grid.addWidget(self.choose_rho_input, 4, 1)
grid.addWidget(self.btn_apply, 5, 0, 1, 2, Qt.AlignCenter)
self.training_dialog.setLayout(grid)
def open_file(self) -> None:
"""
Opens a file dialog used to load image with a mathematical expression
"""
file_name, _ = QFileDialog.getOpenFileName(self, 'Open file')
self.browsed_file.setText(file_name)
self.file_path = file_name
self.w = min(530, QPixmap(file_name).width())
self.h = min(270, QPixmap(file_name).height())
self.label_image.setPixmap(
QPixmap(file_name).scaled(self.w, self.h, Qt.KeepAspectRatio))
self.label_state.setText("Loaded formula from file " + file_name)
self.btn_solve.setEnabled(True)
def save_weights(self) -> None:
"""
Opens a file dialog used to save network parameters to a .csv file
"""
file_name, _ = QFileDialog.getSaveFileName(self,
'Save file',
filter="csv(*.csv)")
if file_name != '':
file = open(file_name, 'w')
with open(file_name):
self.network.save(file_name)
file.close()
self.label_state.setText("Saving weights to file " + file_name)
def load_weights(self) -> None:
"""
Opens a file dialog used to load network parameters from a .csv file
"""
file_name, _ = QFileDialog.getOpenFileName(self,
'Open file',
filter="csv(*.csv)")
if file_name != '':
self.network.load(file_name)
self.label_state.setText("Loaded weights from file " + file_name)
self.network_ready = True
def solve_formula(self) -> None:
"""
Solves the formula recognized in the image by the neural network
"""
if self.network_ready:
self.label_state.setText(
"Evaluating formula using neural network.")
self.formula.load(self.file_path)
self.solved_formula.setText(self.formula.evaluate(self.network))
self.label_image.setPixmap(
QPixmap("data/_output.png").scaled(self.w, self.h,
Qt.KeepAspectRatio))
else:
self.label_state.setText(
"Please train network or load weights first.")
def open_training_dialog(self) -> None:
"""
Opens the dialog before training in which the user sets the number of iterations and additional parameters.
"""
self.training_dialog.exec()
def get_training_params(self) -> None:
"""
Retrieves the training parameters from closed dialog and starts training with
"""
self.iterations = int(self.choose_iterations_input.text())
self.ro = float(self.choose_rho_input.text())
self.alpha = float(self.choose_alpha_input.text())
self.training_dialog.close()
t = threading.Thread(target=self.network.fit,
args=(self.iterations, self.fonts.data,
self.fonts.labels, self.ro, self.alpha,
self.set_progress_bar))
t.start()
self.label_state.setText("Training network for " +
str(self.iterations) + " iterations and " +
str(len(self.fonts.data)) + " samples.")
self.network_ready = True
def set_progress_bar(self, progress) -> None:
"""
Updates the training progress bar with data from neural network
@param progress: Integer in range from 0 to 100 representing training progress in percentage
"""
self.loading.setValue(round(float(progress)))
value = self.loading.value()
if value < 100:
self.btn_train.setEnabled(False)
self.btn_load_weights.setEnabled(False)
self.btn_save_weights.setEnabled(False)
self.btn_browse.setEnabled(False)
else:
self.label_state.setText("Network successfully trained.")
self.btn_train.setEnabled(True)
self.btn_load_weights.setEnabled(True)
self.btn_save_weights.setEnabled(True)
self.btn_browse.setEnabled(True)
