ok-script 0.0.16__tar.gz → 0.0.18__tar.gz

{ok-script-0.0.16 → ok-script-0.0.18}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ok-script
-Version: 0.0.16
+Version: 0.0.18
 Summary: Automation with Computer Vision for Python
 Home-page: https://github.com/ok-oldking/ok-script
 Author: ok-oldking

{ok-script-0.0.16 → ok-script-0.0.18}/ok/OK.py

@@ -82,7 +82,7 @@ class OK:
     def do_init(self):
         logger.info(f"initializing {self.__class__.__name__}, config: {self.config}")
         if self.config.get('ocr'):
-            from rapidocr_onnxruntime import RapidOCR
+            from rapidocr_openvino import RapidOCR
             self.ocr = RapidOCR()
 
         config_logger(self.config)

{ok-script-0.0.16 → ok-script-0.0.18}/ok/feature/Box.py

@@ -91,14 +91,14 @@ class Box:
         return self.x + self.width / 2, self.y + self.height / 2
 
     def find_closest_box(self, direction: str, boxes: list, condition=None):
-        orig_center_x, orig_center_y = self.center()
+        orig_x, orig_y, orig_w, orig_h = self.x, self.y, self.width, self.height
 
         def distance_criteria(box):
-            # Calculate center points for comparison
-            box_center_x, box_center_y = box.center()
+            box_x, box_y, box_w, box_h = box.x, box.y, box.width, box.height
+
+            dx = max(orig_x - (box_x + box_w), box_x - (orig_x + orig_w), 0)
+            dy = max(orig_y - (box_y + box_h), box_y - (orig_y + orig_h), 0)
 
-            dx = box_center_x - orig_center_x
-            dy = box_center_y - orig_center_y
             distance = math.sqrt(dx ** 2 + dy ** 2)
             if box == self:
                 distance = float('inf')
@@ -110,6 +110,8 @@ class Box:
                 pass
             elif direction == 'right' and box.x - (self.x + self.width / 2) >= 0:
                 pass
+            elif direction == 'all':
+                pass
             else:
                 distance = float('inf')
             return check_condition(box, distance)
@@ -121,7 +123,6 @@ class Box:
                 return float('inf')
 
         filtered_boxes = sorted(boxes, key=distance_criteria)
-        # Removed debug print statement for cleanliness
 
         for box in filtered_boxes:
             if distance_criteria(box) != float('inf'):

{ok-script-0.0.16 → ok-script-0.0.18}/ok/gui/debug/Screenshot.py

@@ -13,7 +13,7 @@ import ok.gui
 from ok.feature.Box import Box
 from ok.gui.Communicate import communicate
 from ok.logging.Logger import get_logger
-from ok.util.path import get_path_relative_to_exe, find_first_existing_file, clear_folder
+from ok.util.path import get_path_relative_to_exe, find_first_existing_file, clear_folder, sanitize_filename
 
 logger = get_logger(__name__)
 
@@ -130,10 +130,15 @@ class Screenshot(QObject):
 
         # Convert to string with milliseconds
         time_string = now.strftime("%H_%M_%S_%f")
+        file_name = time_string
         if name:
-            time_string += f"_{name}"
-        file = os.path.join(folder, f"{time_string}.png")
-        pil_image.save(file)
+            name = sanitize_filename(name)
+            file_name = f"{time_string}_{name}"
+        file = os.path.join(folder, f"{file_name}.png")
+        try:
+            pil_image.save(file)
+        except OSError:
+            pil_image.save(os.path.join(folder, f"{time_string}.png"))
 
     def stop(self):
         if self.task_queue is not None:

{ok-script-0.0.16 → ok-script-0.0.18}/ok/gui/tasks/TaskCard.py

@@ -116,7 +116,7 @@ class TaskButtons(QWidget):
                 self.start_button.setText(self.tr("Resume"))
                 self.start_button.show()
                 self.pause_button.hide()
-                self.stop_button.hide()
+                self.stop_button.show()
             elif self.task.running:
                 self.start_button.hide()
                 self.stop_button.show()

{ok-script-0.0.16 → ok-script-0.0.18}/ok/ocr/OCR.py

@@ -1,5 +1,7 @@
 import time
 
+import cv2
+
 from ok.feature.Box import Box, sort_boxes, find_boxes_by_name
 from ok.gui.Communicate import communicate
 from ok.logging.Logger import get_logger
@@ -9,22 +11,31 @@ logger = get_logger(__name__)
 
 class OCR:
     executor = None
-    default_threshold = 0.6
+    ocr_default_threshold = 0.5
+    ocr_target_height = 0
 
-    def ocr(self, box: Box = None, match=None, threshold=0):
+    def ocr(self, box: Box = None, match=None, threshold=0, frame=None, target_height=0):
         if self.paused:
             self.sleep(1)
         if threshold == 0:
-            threshold = self.default_threshold
+            threshold = self.ocr_default_threshold
+        if target_height == 0:
+            target_height = self.ocr_target_height
         start = time.time()
-        image = self.frame
+        if frame is not None:
+            image = frame
+        else:
+            image = self.frame
         if image is None:
             raise Exception("ocr no frame")
         else:
+            original_height = image.shape[0]
             if box is not None:
                 x, y, w, h = box.x, box.y, box.width, box.height
                 image = image[y:y + h, x:x + w]
 
+            image, scale_factor = resize_image(image, original_height, target_height)
+
             result, _ = self.executor.ocr(image, use_det=True, use_cls=False, use_rec=True)
             detected_boxes = []
             # Process the results and create Box objects
@@ -37,18 +48,43 @@ class OCR:
                         detected_box = Box(int(pos[0][0]), int(pos[0][1]), int(pos[2][0] - pos[0][0]),
                                            int(pos[2][1] - pos[0][1]),
                                            confidence, text)
+                        scale_box(detected_box, scale_factor)
                         if box is not None:
                             detected_box.x += box.x
                             detected_box.y += box.y
                         detected_boxes.append(detected_box)
                 if match is not None:
                     detected_boxes = find_boxes_by_name(detected_boxes, match)
+
             communicate.emit_draw_box("ocr", detected_boxes, "red")
             communicate.emit_draw_box("ocr_zone", box, "blue")
-            logger.debug(f"ocr_zone {box} found result: {len(detected_boxes)}) time: {time.time() - start}")
+            logger.debug(
+                f"ocr_zone {box} found result: {len(detected_boxes)}) time: {(time.time() - start):.2f} scale_factor: {scale_factor:.2f}")
             return sort_boxes(detected_boxes)
 
     def find_text(self, text, box: Box = None, confidence=0):
         for result in self.ocr(box, confidence):
             if result.name == text:
                 return result
+
+
+def resize_image(image, original_height, target_height):
+    scale_factor = 1
+    if target_height > 0 and original_height >= 2 * target_height:
+        image_height, image_width = image.shape[:2]
+        times = int(original_height / target_height)
+        scale_factor = 1 / times
+        # Calculate the new width to maintain the aspect ratio
+        new_width = round(image_width * scale_factor)
+        new_height = round(image_height * scale_factor)
+        # Resize the image
+        image = cv2.resize(image, (new_width, new_height))
+    return image, scale_factor
+
+
+def scale_box(box, scale_factor):
+    if scale_factor != 1:
+        box.x = round(box.x / scale_factor)
+        box.y = round(box.y / scale_factor)
+        box.width = round(box.width / scale_factor)
+        box.height = round(box.height / scale_factor)

{ok-script-0.0.16 → ok-script-0.0.18}/ok/task/ExecutorOperation.py

@@ -5,6 +5,9 @@ from ok.gui.Communicate import communicate
 class ExecutorOperation:
     executor = None
 
+    def exit_is_set(self):
+        return self.executor.exit_event.is_set()
+
     def box_in_horizontal_center(self, box, off_percent=0.02):
         if box is None:
             return False

{ok-script-0.0.16 → ok-script-0.0.18}/ok_script.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ok-script
-Version: 0.0.16
+Version: 0.0.18
 Summary: Automation with Computer Vision for Python
 Home-page: https://github.com/ok-oldking/ok-script
 Author: ok-oldking

{ok-script-0.0.16 → ok-script-0.0.18}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r", encoding="utf-8") as fh:
 
 setuptools.setup(
     name="ok-script",
-    version="0.0.16",
+    version="0.0.18",
     author="ok-oldking",
     author_email="firedcto@gmail.com",
     description="Automation with Computer Vision for Python",