def add_datasets_combobox(self): self.dataset_id = Iris self.datasets_combobox = ZLabeledComboBox(self, "Datasets", Qt.Horizontal) # We use the following datasets of sklearn to test XGBClassifier. self.datasets = { "Iris": Iris, "Digits": Digits, "Wine": Wine, "BreastCancer": BreastCancer } title = self.get_title() self.setWindowTitle("Iris" + " - " + title) self.datasets_combobox.add_items(self.datasets.keys()) self.datasets_combobox.add_activated_callback(self.datasets_activated) self.datasets_combobox.set_current_text(self.dataset_id) self.start_button = ZPushButton("Start", self) self.clear_button = ZPushButton("Clear", self) self.start_button.add_activated_callback(self.start_button_activated) self.clear_button.add_activated_callback(self.clear_button_activated) self.datasets_combobox.add(self.start_button) self.datasets_combobox.add(self.clear_button) self.set_top_dock(self.datasets_combobox)
def add_control_pane(self, fixed_width=220): # Control pane widget self.vpane = ZVerticalPane(self, fixed_width) self.block_size = 11 self.adaptive_method_id = 0; self.threshold_type_id = 0; self.methods = {"ADAPTIVE_THRESH_MEAN_C": cv2.ADAPTIVE_THRESH_MEAN_C, "ADAPTIVE_THRESH_GAUSSIAN_C": cv2.ADAPTIVE_THRESH_GAUSSIAN_C} self.types = {"THRESH_BINARY": cv2.THRESH_BINARY , "THRESH_BINARY_INV": cv2.THRESH_BINARY_INV } self.adaptive_method = ZLabeledComboBox(self.vpane, "AdaptiveMethod") self.adaptive_method.add_items(list(self.methods.keys() )) self.adaptive_method.add_activated_callback(self.adaptive_method_activated) self.threshold_type = ZLabeledComboBox(self.vpane, "ThresholdType") self.threshold_type.add_items(list(self.types.keys()) ) self.threshold_type.add_activated_callback(self.threshold_type_activated) self.labeled_slider = ZLabeledSlider(self.vpane, "BlockSize", take_odd =True, minimum=3, maximum=43, value=self.block_size, fixed_width=200) self.labeled_slider.add_value_changed_callback(self.slider_value_changed) self.vpane.add(self.adaptive_method) self.vpane.add(self.threshold_type) self.vpane.add(self.labeled_slider) self.set_right_dock(self.vpane)
def add_datasets_combobox(self): self.dataset_id = Boston self.datasets_combobox = ZLabeledComboBox(self, "Datasets", Qt.Horizontal) self.datasets_combobox.setFont(self.font) # We use the following datasets of sklearn to test XGBRegressor. self.datasets = {"Boston": Boston, "Diabetes": Diabetes} title = self.get_title() self.setWindowTitle("Boston" + " - " + title) self.datasets_combobox.add_items(self.datasets.keys()) self.datasets_combobox.add_activated_callback(self.datasets_activated) self.datasets_combobox.set_current_text(self.dataset_id) self.start_button = ZPushButton("Start", self) self.clear_button = ZPushButton("Clear", self) self.start_button.setFont(self.font) self.clear_button.setFont(self.font) self.start_button.add_activated_callback(self.start_button_activated) self.clear_button.add_activated_callback(self.clear_button_activated) self.datasets_combobox.add(self.start_button) self.datasets_combobox.add(self.clear_button) self.set_top_dock(self.datasets_combobox)
def add_datasets_combobox(self): datasetkey = list(self.datasets.keys())[0] self.dataset_id = self.datasets[datasetkey] print("Current combobox item {} {}".format(datasetkey, self.dataset_id)) self.datasets_combobox = ZLabeledComboBox(self, "Datasets", Qt.Horizontal) self.datasets_combobox.setFont(self.font) title = self.get_title() self.setWindowTitle(self.__class__.__name__ + " - " + title) self.datasets_combobox.add_items(self.datasets.keys()) self.datasets_combobox.add_activated_callback(self.datasets_activated) self.datasets_combobox.set_current_text(self.dataset_id) self.classifier_button = ZPushButton("Classify", self) self.classifier_button.setEnabled(False) self.clear_button = ZPushButton("Clear", self) self.classifier_button.add_activated_callback(self.classifier_button_activated) self.clear_button.add_activated_callback(self.clear_button_activated) self.datasets_combobox.add(self.classifier_button) self.datasets_combobox.add(self.clear_button) self.set_top_dock(self.datasets_combobox)
class MainView(ZApplicationView): class BinarizedImageView(ZOpenCVImageView): def __init__(self, parent): ZOpenCVImageView.__init__(self, parent) # Constructor def __init__(self, title, x, y, width, height): super(MainView, self).__init__(title, x, y, width, height) self.raw_video_queue = mp.Queue() self.to_binarize_video_queue = mp.Queue() self.binarized_video_queue = mp.Queue() # Binarize initial settings #self.block_size = 11 #self.adaptive_method_id = 0; #self.threshold_type_id = 0; self.binarize_video_settings_queue = mp.Queue() #binarize_video_settings_queue.put_nowait((self.block_size, \ # self.adaptive_method_id, \ # self.threshold_type_id)) self.time_snapshot_seconds = None # initialized to None . Used for rate limiting processing pipelines self.processing_rate_limit_hz = 20 # 30 Hz seems to be about the maximum if this camera self.raw_image_view = ZOpenCVImageView(self) self.binarized_image_view = self.BinarizedImageView(self) self.add(self.raw_image_view) self.add(self.binarized_image_view) title = name self.setWindowTitle(title) self.show() def add_control_pane(self, fixed_width=220): # Control pane widget self.vpane = ZVerticalPane(self, fixed_width) #""" self.block_size = 11 self.adaptive_method_id = 0 self.threshold_type_id = 0 #""" self.methods = { "ADAPTIVE_THRESH_MEAN_C": cv2.ADAPTIVE_THRESH_MEAN_C, "ADAPTIVE_THRESH_GAUSSIAN_C": cv2.ADAPTIVE_THRESH_GAUSSIAN_C } self.types = { "THRESH_BINARY": cv2.THRESH_BINARY, "THRESH_BINARY_INV": cv2.THRESH_BINARY_INV } self.adaptive_method = ZLabeledComboBox(self.vpane, "AdaptiveMethod") self.adaptive_method.add_items(list(self.methods.keys())) self.adaptive_method.add_activated_callback( self.adaptive_method_activated) self.threshold_type = ZLabeledComboBox(self.vpane, "ThresholdType") self.threshold_type.add_items(list(self.types.keys())) self.threshold_type.add_activated_callback( self.threshold_type_activated) self.labeled_slider = ZLabeledSlider(self.vpane, "BlockSize", take_odd=True, minimum=3, maximum=43, value=self.block_size, fixed_width=200) self.labeled_slider.add_value_changed_callback( self.slider_value_changed) self.vpane.add(self.adaptive_method) self.vpane.add(self.threshold_type) self.vpane.add(self.labeled_slider) self.set_right_dock(self.vpane) def slider_value_changed(self, value): self.block_size = int(value) if self.block_size % 2 == 0: # Block size should be odd. self.block_size = int((self.block_size * 2) / 2 + 1) #print("slider_value_changed:{}".format(block_size)) #self.binarized_image_view.binarize(self.adaptive_method_id, self.threshold_type_id, self.block_size) settings = (self.adaptive_method_id, self.threshold_type_id, self.block_size) self.binarize_video_settings_queue.put_nowait(settings) def adaptive_method_activated(self, text): #print("adaptive_method_activated:{}".format(text)) self.adaptive_method_id = self.methods[text] #self.binarized_image_view.binarize(self.adaptive_method_id, self.threshold_type_id, self.block_size) settings = (self.adaptive_method_id, self.threshold_type_id, self.block_size) self.binarize_video_settings_queue.put_nowait(settings) def threshold_type_activated(self, text): #print("threshold_type_activated:{}".format(text)) self.threshold_type_id = self.types[text] #self.binarized_image_view.binarize(self.adaptive_method_id, self.threshold_type_id, self.block_size) settings = (self.adaptive_method_id, self.threshold_type_id, self.block_size) self.binarize_video_settings_queue.put_nowait(settings) # Read a frame from a video buffer of the video capture, and set it the image view to draw it on the imag view def render(self): return_true = False #return_true = True #""" if not self.raw_video_queue.empty(): raw_frame = self.raw_video_queue.get_nowait() if raw_frame.any() != None: self.raw_image_view.set_opencv_image(raw_frame) self.raw_image_view.update() # Send frame to processing queue # rate limited current_time_seconds = time() if self.time_snapshot_seconds == None: self.time_snapshot_seconds = current_time_seconds if current_time_seconds > self.time_snapshot_seconds: # Avoid dividing by zero #print(1/(current_time_seconds-self.time_snapshot_seconds)) if ((1 / (current_time_seconds - self.time_snapshot_seconds)) <= self.processing_rate_limit_hz): # Rate limit #print("Sending raw frame for further processing!") self.to_binarize_video_queue.put_nowait(raw_frame) self.time_snapshot_seconds = current_time_seconds return_true = True if not self.binarized_video_queue.empty(): binarized_frame = self.binarized_video_queue.get_nowait() if binarized_frame.any() != None: self.binarized_image_view.set_opencv_image(binarized_frame) self.binarized_image_view.update() return_true = True #""" if return_true: return True else: return False def file_quit(self): self.terminated = True def closeEvent(self, ce): self.terminated = True
class MainView(ZApplicationView): # Class variables # ClassifierView Constructor def __init__(self, title, x, y, width, height): super(MainView, self).__init__(title, x, y, width, height) self.font = QFont("Arial", 10) self.setFont(self.font) # 1 Add a labeled combobox to top dock area self.add_datasets_combobox() # 2 Add a textedit to the left pane of the center area. self.text_editor = QTextEdit() self.text_editor.setLineWrapColumnOrWidth(600) self.text_editor.setLineWrapMode(QTextEdit.FixedPixelWidth) # 3 Add a description to display dataset.DESCR. self.description = QTextEdit() self.description.setLineWrapColumnOrWidth(600) self.description.setLineWrapMode(QTextEdit.FixedPixelWidth) # 4 Add a tabbed_window to the tabbed_window. self.tabbed_window = ZTabbedWindow(self, 0, 0, width / 2, height) # 5 Add a figure_view to the tabbed_window. self.figure_view = ZScalableScrolledFigureView(self, 0, 0, width / 2, height) self.add(self.text_editor) self.add(self.tabbed_window) self.tabbed_window.add("Description", self.description) self.tabbed_window.add("ConfusionMatrix", self.figure_view) self.figure_view.hide() self.show() def add_datasets_combobox(self): self.dataset_id = Iris self.datasets_combobox = ZLabeledComboBox(self, "Datasets", Qt.Horizontal) # We use the following datasets of sklearn to test XGBClassifier. self.datasets = { "Iris": Iris, "Digits": Digits, "Wine": Wine, "BreastCancer": BreastCancer } title = self.get_title() self.setWindowTitle("Iris" + " - " + title) self.datasets_combobox.add_items(self.datasets.keys()) self.datasets_combobox.add_activated_callback(self.datasets_activated) self.datasets_combobox.set_current_text(self.dataset_id) self.start_button = ZPushButton("Start", self) self.clear_button = ZPushButton("Clear", self) self.start_button.add_activated_callback(self.start_button_activated) self.clear_button.add_activated_callback(self.clear_button_activated) self.datasets_combobox.add(self.start_button) self.datasets_combobox.add(self.clear_button) self.set_top_dock(self.datasets_combobox) def write(self, text): self.text_editor.append(text) self.text_editor.repaint() def datasets_activated(self, text): self.dataset_id = self.datasets[text] title = self.get_title() self.setWindowTitle(text + " - " + title) def start_button_activated(self, text): self.model = LightGBMClassifierModel(self.dataset_id, self) self.start_button.setEnabled(False) self.clear_button.setEnabled(False) try: self.model.run() except: pass self.start_button.setEnabled(True) self.clear_button.setEnabled(True) def clear_button_activated(self, text): self.text_editor.setText("") self.description.setText("") self.figure_view.hide() plt.close() def visualize(self, cmatrix): self.figure_view.show() plt.close() sns.set() df = pd.DataFrame(cmatrix) sns.heatmap(df, annot=True, fmt="d") # Set a new figure to the figure_view. self.figure_view.set_figure(plt)
class MainView(ZApplicationView): # MainView Constructor def __init__(self, title, x, y, width, height): super(MainView, self).__init__(title, x, y, width, height) self.font = QFont("Arial", 10) self.setFont(self.font) # 1 Add a labeled combobox to top dock area self.add_datasets_combobox() # 2 Add a textedit to the left pane of the center area. self.text_editor = QTextEdit() self.text_editor.setFont(self.font) self.text_editor.setLineWrapColumnOrWidth(600) self.text_editor.setLineWrapMode(QTextEdit.FixedPixelWidth) self.description = QTextEdit() self.description.setFont(self.font) self.description.setLineWrapColumnOrWidth(600) self.description.setLineWrapMode(QTextEdit.FixedPixelWidth) # 3 Add a tabbled_window to the right pane of the center area. self.tabbed_window = ZTabbedWindow(self, 0, 0, width/2, height) # 4 Add a figure_view to the tabbed_window self.figure_view = ZScalableScrolledFigureView(self, 0, 0, width/2, height) self.add(self.text_editor) self.add(self.tabbed_window) self.tabbed_window.add("Description", self.description) self.tabbed_window.add("Importances", self.figure_view) self.figure_view.hide() self.show() def add_datasets_combobox(self): self.dataset_id = Boston self.datasets_combobox = ZLabeledComboBox(self, "Datasets", Qt.Horizontal) # We use the following datasets of sklearn to test RandomForestRegressor. self.datasets = {"Boston": Boston, "Diabetes": Diabetes} title = self.get_title() self.setWindowTitle( "Boston" + " - " + title) self.datasets_combobox.add_items(self.datasets.keys()) self.datasets_combobox.add_activated_callback(self.datasets_activated) self.datasets_combobox.set_current_text(self.dataset_id) self.start_button = ZPushButton("Start", self) self.clear_button = ZPushButton("Clear", self) self.start_button.add_activated_callback(self.start_button_activated) self.clear_button.add_activated_callback(self.clear_button_activated) self.datasets_combobox.add(self.start_button) self.datasets_combobox.add(self.clear_button) self.set_top_dock(self.datasets_combobox) def write(self, text): self.text_editor.append(text) self.text_editor.repaint() def datasets_activated(self, text): self.dataset_id = self.datasets[text] title = self.get_title() self.setWindowTitle(text + " - " + title) def start_button_activated(self, text): self.model = RandomForestRegressorModel(self.dataset_id, self) self.start_button.setEnabled(False) self.clear_button.setEnabled(False) try: self.model.run() except: pass self.start_button.setEnabled(True) self.clear_button.setEnabled(True) def file_new(self): self.text_editor.setText("") self.description.setText("") self.figure_view.hide() plt.close() def clear_button_activated(self, text): self.file_new() def visualize(self, importances): self.figure_view.show() plt.close() plt.title("Importances in Regression Model") importances.plot(kind = "barh") self.figure_view.set_figure( plt)
class ZImageClassifierView(ZApplicationView): # Class variables # ClassifierView Constructor def __init__(self, title, x, y, width, height, datasets={"ImageModel": 0}): super(ZImageClassifierView, self).__init__(title, x, y, width, height) self.font = QFont("Arial", 10) self.setFont(self.font) self.datasets = datasets self.model_loaded = False self.class_names_set = [None, None] # Image filename to be classified self.filename = None # Target image to be classified self.image = None # 1 Add a labeled combobox to top dock area self.add_datasets_combobox() # 2 Add a textedit to the left pane of the center area. self.text_editor = QTextEdit() self.text_editor.setLineWrapColumnOrWidth(600) self.text_editor.setLineWrapMode(QTextEdit.FixedPixelWidth) self.text_editor.setGeometry(0, 0, width/2, height) # 3 Add a tabbed_window the rigth pane of the center area. self.tabbed_window = ZTabbedWindow(self, 0, 0, width/2, height) # 4 Add a imageview to the tabbed_window. self.image_view = ZScalableScrolledImageView(self, 0, 0, width/3, height/3) self.tabbed_window.add("SourceImage", self.image_view) # 5 Add a test_imageview to the right pane of the center area. self.test_image_view = ZScalableScrolledImageView(self, 0, 0, width/3, height/3) self.tabbed_window.add("TestImage", self.test_image_view) self.add(self.text_editor) self.add(self.tabbed_window) def add_datasets_combobox(self): datasetkey = list(self.datasets.keys())[0] self.dataset_id = self.datasets[datasetkey] print("Current combobox item {} {}".format(datasetkey, self.dataset_id)) self.datasets_combobox = ZLabeledComboBox(self, "Datasets", Qt.Horizontal) self.datasets_combobox.setFont(self.font) title = self.get_title() self.setWindowTitle(self.__class__.__name__ + " - " + title) self.datasets_combobox.add_items(self.datasets.keys()) self.datasets_combobox.add_activated_callback(self.datasets_activated) self.datasets_combobox.set_current_text(self.dataset_id) self.classifier_button = ZPushButton("Classify", self) self.classifier_button.setEnabled(False) self.clear_button = ZPushButton("Clear", self) self.classifier_button.add_activated_callback(self.classifier_button_activated) self.clear_button.add_activated_callback(self.clear_button_activated) self.datasets_combobox.add(self.classifier_button) self.datasets_combobox.add(self.clear_button) self.set_top_dock(self.datasets_combobox) def write(self, text): self.text_editor.append(text) self.text_editor.repaint() def datasets_activated(self, text): pass # Show FileOpenDialog and select an image file. def file_open(self): if self.model_loaded: options = QFileDialog.Options() filename, _ = QFileDialog.getOpenFileName(self,"FileOpenDialog", "", "All Files (*);;Image Files (*.png;*jpg;*.jpeg)", options=options) if filename: self.filename = filename self.load_file(filename) self.classifier_button.setEnabled(True) else: QMessageBox.warning(self, "ImageClassifier: Weight File Missing", "Please run: python RoadSignsModel.py " + str(self.dataset_id)) def remove_alpha_channel(self, array): shape = array.shape if len(shape) ==3: w, h, c = shape if c == 4: #(w, h, 4) -> (w, h, 3) #print("Remove the alpha channel of array") array = array[:,:,:3] return array def load_file(self, filename): from keras.preprocessing.image import load_img, img_to_array try: image_cropper = ZPILImageCropper(filename) # 1 Crop larget_central square region from an image of filename. cropped_image = image_cropper.crop_largest_central_square_region() # 2 Load an image from the cropped_fle. self.image_view.set_image(img_to_array(cropped_image)) self.image_view.update_scale() self.set_filenamed_title(filename) # 3 Resize the cropped_image self.image = cropped_image.resize(self.image_size) # 4 Convert the self.image to numpy ndarray and remove alpha channel. self.image = self.remove_alpha_channel(img_to_array(self.image)) # 5 Set self.nadarryy to the test_image_view. self.test_image_view.set_image(self.image) # 6 Convert self.image in range[0-1.0] self.image = self.image.astype('float32')/255.0 # 7 Expand the dimension of the self.image self.image = np.expand_dims(self.image, axis=0) #print(self.image.shape) except: self.write(formatted_traceback()) def classifier_button_activated(self, text): self.classifier_button.setEnabled(False) self.clear_button.setEnabled(False) try: self.classify() except: self.write(formatted_traceback()) self.classifier_button.setEnabled(True) self.clear_button.setEnabled(True) def get_top_five(self, predictions, classes, K=5): pred = predictions[0] indices = np.argpartition(pred, -K)[-K:] results = [] for i in indices: results.append([pred[i], classes[i]]) results = sorted(results, reverse=True) return results def get_class_names(self, path="./class_names.txt"): classes = [] with open(path, "r") as file: classes = [s.strip() for s in file.readlines()] return classes def classify(self): self.write("------------------------------------------------------------") self.write("classify start") self.write(self.filename) prediction = self.model.predict(self.image) pred = np.argmax(prediction, axis=1) #self.write("Prediction: index " + str(pred)) class_names = self.model.classes if pred >0 or pred <len(class_names): self.write("Prediction:" + class_names[int(pred)]) self.write("classify end") def clear_button_activated(self, text): self.text_editor.setText("") pass
class MainView(ZApplicationView): class BinarizedImageView(ZOpenCVImageView): def __init__(self, parent): ZOpenCVImageView.__init__(self, parent) def load_source_image(self, source_image): image = cv2.GaussianBlur(source_image, (3,3), 0, 0, borderType = cv2.BORDER_DEFAULT ); #image = cv2.GaussianBlur(source_image, (3,3), 0, 0) self.gray_image = cv2.cvtColor(source_image, cv2.COLOR_RGB2GRAY) def binarize(self, adaptive_method_id, threshold_type_id, block_size): MAX_PIXEL_VALUE = 255 C = 9.0 binarized_image = cv2.adaptiveThreshold(self.gray_image, MAX_PIXEL_VALUE, adaptive_method_id, threshold_type_id, block_size, C); #return binarized_image self.set_opencv_image(binarized_image) self.update() # Constructor def __init__(self, title, x, y, width, height): super(MainView, self).__init__(title, x, y, width, height) self.raw_video_queue = mp.Queue() self.raw_image_view = ZOpenCVImageView(self) self.binarized_image_view = self.BinarizedImageView(self) self.add(self.raw_image_view) self.add(self.binarized_image_view) title = name self.setWindowTitle(title) self.show() def add_control_pane(self, fixed_width=220): # Control pane widget self.vpane = ZVerticalPane(self, fixed_width) self.block_size = 11 self.adaptive_method_id = 0; self.threshold_type_id = 0; self.methods = {"ADAPTIVE_THRESH_MEAN_C": cv2.ADAPTIVE_THRESH_MEAN_C, "ADAPTIVE_THRESH_GAUSSIAN_C": cv2.ADAPTIVE_THRESH_GAUSSIAN_C} self.types = {"THRESH_BINARY": cv2.THRESH_BINARY , "THRESH_BINARY_INV": cv2.THRESH_BINARY_INV } self.adaptive_method = ZLabeledComboBox(self.vpane, "AdaptiveMethod") self.adaptive_method.add_items(list(self.methods.keys() )) self.adaptive_method.add_activated_callback(self.adaptive_method_activated) self.threshold_type = ZLabeledComboBox(self.vpane, "ThresholdType") self.threshold_type.add_items(list(self.types.keys()) ) self.threshold_type.add_activated_callback(self.threshold_type_activated) self.labeled_slider = ZLabeledSlider(self.vpane, "BlockSize", take_odd =True, minimum=3, maximum=43, value=self.block_size, fixed_width=200) self.labeled_slider.add_value_changed_callback(self.slider_value_changed) self.vpane.add(self.adaptive_method) self.vpane.add(self.threshold_type) self.vpane.add(self.labeled_slider) self.set_right_dock(self.vpane) def slider_value_changed(self, value): self.block_size = int(value) if self.block_size % 2 == 0: # Block size should be odd. self.block_size = int((self.block_size * 2)/2 + 1) #print("slider_value_changed:{}".format(block_size)) self.binarized_image_view.binarize(self.adaptive_method_id, self.threshold_type_id, self.block_size) def adaptive_method_activated(self, text): #print("adaptive_method_activated:{}".format(text)) self.adaptive_method_id = self.methods[text] self.binarized_image_view.binarize(self.adaptive_method_id, self.threshold_type_id, self.block_size) def threshold_type_activated(self, text): #print("threshold_type_activated:{}".format(text)) self.threshold_type_id = self.types[text] self.binarized_image_view.binarize(self.adaptive_method_id, self.threshold_type_id, self.block_size) # Read a frame from a video buffer of the video capture, and set it the image view to draw it on the imag view def render(self): #return_true = False if not self.raw_video_queue.empty(): raw_frame = self.raw_video_queue.get_nowait() if raw_frame.any() != None: self.raw_image_view.set_opencv_image(raw_frame) self.raw_image_view.update() self.binarized_image_view.load_source_image(raw_frame) self.binarized_image_view.binarize(self.adaptive_method_id, self.threshold_type_id, self.block_size) return True """ if not self.binarized_video_queue.empty(): binarized_frame = self.binarized_video_queue.get_notwait() if binarized_frame.any() != None: self.binarized_image_view.set_opencv_image(binarized_frame) self.binarized_image_view.update() return_true = True if return_true: return True else: return False """ return False """ def load_file(self, filename): self.image_view_3.load(filename) self.set_filenamed_title(filename) """ def file_quit(self): self.terminated = True def closeEvent(self, ce): self.terminated = True