coralnet-toolbox 0.0.72__py2.py3-none-any.whl → 0.0.74__py2.py3-none-any.whl

coralnet_toolbox/MachineLearning/ExportDataset/QtDetect.py

@@ -53,6 +53,10 @@ class Detect(Base):
         self.include_rectangles_checkbox.setEnabled(True)  # Enable user to uncheck rectangles if desired
         self.include_polygons_checkbox.setChecked(True)
         self.include_polygons_checkbox.setEnabled(True)  # Already enabled
+
+        # Explicitly enable negative sample options for detection
+        self.include_negatives_radio.setEnabled(True)
+        self.exclude_negatives_radio.setEnabled(True)
         
     def create_dataset(self, output_dir_path):
         """
@@ -106,12 +110,20 @@ class Detect(Base):
         Process and save detection annotations.
 
         Args:
-            annotations (list): List of annotations.
+            annotations (list): List of annotations for this split.
             split_dir (str): Path to the split directory.
             split (str): Split name (e.g., "Training", "Validation", "Testing").
         """
-        # Get unique image paths
-        image_paths = list(set(a.image_path for a in annotations))
+        # Determine the full list of images for this split (including negatives)
+        if split == "Training":
+            image_paths = self.train_images
+        elif split == "Validation":
+            image_paths = self.val_images
+        elif split == "Testing":
+            image_paths = self.test_images
+        else:
+            image_paths = []
+
         if not image_paths:
             return
 
@@ -124,6 +136,7 @@ class Detect(Base):
         for image_path in image_paths:
             yolo_annotations = []
             image_height, image_width = rasterio_open(image_path).shape
+            # Filter the annotations passed to this function to get only those for the current image
             image_annotations = [a for a in annotations if a.image_path == image_path]
 
             for image_annotation in image_annotations:
@@ -132,11 +145,11 @@ class Detect(Base):
                 yolo_annotations.append(f"{class_number} {annotation}")
 
             # Save the annotations to a text file
-            file_ext = image_path.split(".")[-1]
-            text_file = os.path.basename(image_path).replace(f".{file_ext}", ".txt")
+            file_ext = os.path.splitext(image_path)[1]
+            text_file = os.path.basename(image_path).replace(file_ext, ".txt")
             text_path = os.path.join(f"{split_dir}/labels", text_file)
 
-            # Write the annotations to the text file
+            # Write the annotations to the text file (creates an empty file for negatives)
             with open(text_path, 'w') as f:
                 for annotation in yolo_annotations:
                     f.write(annotation + '\n')

coralnet_toolbox/MachineLearning/ExportDataset/QtSegment.py

@@ -53,6 +53,10 @@ class Segment(Base):
         self.include_rectangles_checkbox.setEnabled(True)  # Enable rectangles for segmentation
         self.include_polygons_checkbox.setChecked(True)
         self.include_polygons_checkbox.setEnabled(True)  # Enable user to uncheck polygons if desired
+        
+        # Explicitly enable negative sample options for segmentation
+        self.include_negatives_radio.setEnabled(True)
+        self.exclude_negatives_radio.setEnabled(True)
 
     def create_dataset(self, output_dir_path):
         """
@@ -106,12 +110,20 @@ class Segment(Base):
         Process and save segmentation annotations.
 
         Args:
-            annotations (list): List of annotations.
+            annotations (list): List of annotations for this split.
             split_dir (str): Path to the split directory.
             split (str): Split name (e.g., "Training", "Validation", "Testing").
         """
-        # Get unique image paths
-        image_paths = list(set(a.image_path for a in annotations))
+        # Determine the full list of images for this split (including negatives)
+        if split == "Training":
+            image_paths = self.train_images
+        elif split == "Validation":
+            image_paths = self.val_images
+        elif split == "Testing":
+            image_paths = self.test_images
+        else:
+            image_paths = []
+
         if not image_paths:
             return
 
@@ -124,6 +136,7 @@ class Segment(Base):
         for image_path in image_paths:
             yolo_annotations = []
             image_height, image_width = rasterio_open(image_path).shape
+            # Filter the annotations passed to this function to get only those for the current image
             image_annotations = [a for a in annotations if a.image_path == image_path]
 
             for image_annotation in image_annotations:
@@ -132,11 +145,11 @@ class Segment(Base):
                 yolo_annotations.append(f"{class_number} {annotation}")
 
             # Save the annotations to a text file
-            file_ext = image_path.split(".")[-1]
-            text_file = os.path.basename(image_path).replace(f".{file_ext}", ".txt")
+            file_ext = os.path.splitext(image_path)[1]
+            text_file = os.path.basename(image_path).replace(file_ext, ".txt")
             text_path = os.path.join(f"{split_dir}/labels", text_file)
 
-            # Write the annotations to the text file
+            # Write the annotations to the text file (creates an empty file for negatives)
             with open(text_path, 'w') as f:
                 for annotation in yolo_annotations:
                     f.write(annotation + '\n')
@@ -146,7 +159,7 @@ class Segment(Base):
 
             progress_bar.update_progress()
 
-        # Make cursor normal
+        # Reset cursor
         QApplication.restoreOverrideCursor()
         progress_bar.stop_progress()
-        progress_bar.close()
+        progress_bar.close()

coralnet_toolbox/MachineLearning/ImportDataset/QtBase.py

@@ -35,7 +35,7 @@ class DatasetProcessor(QObject):
     """
     status_changed = pyqtSignal(str, int)
     progress_updated = pyqtSignal(int)
-    processing_complete = pyqtSignal(list, list)
+    processing_complete = pyqtSignal(list, list, list)
     error = pyqtSignal(str)
     finished = pyqtSignal()
 
@@ -47,6 +47,7 @@ class DatasetProcessor(QObject):
         self.import_as = import_as  # 'rectangle' or 'polygon' (target format)
         self.rename_on_conflict = rename_on_conflict
         self.is_running = True
+        self.parsing_errors = []  # To collect errors instead of printing
 
     def stop(self):
         self.is_running = False
@@ -81,7 +82,7 @@ class DatasetProcessor(QObject):
 
             # Step 4: Emit results for GUI to consume
             image_paths = list(image_label_paths.keys())
-            self.processing_complete.emit(raw_annotations, image_paths)
+            self.processing_complete.emit(raw_annotations, image_paths, self.parsing_errors)
 
         except Exception as e:
             # Catch-all for any error during processing
@@ -146,7 +147,7 @@ class DatasetProcessor(QObject):
             with open(label_path, 'r') as file:
                 lines = file.readlines()
 
-            for line in lines:
+            for line_num, line in enumerate(lines):
                 try:
                     parts = list(map(float, line.split()))
                     class_id = int(parts[0])
@@ -195,8 +196,11 @@ class DatasetProcessor(QObject):
 
                     all_raw_annotations.append(raw_ann_data)
                 except (ValueError, IndexError) as e:
-                    # Skip malformed lines and print a warning
-                    print(f"Skipping malformed line in {label_path}: {line.strip()} ({e})")
+                    # Log the malformed line error instead of printing
+                    error_msg = (f"In file '{os.path.basename(label_path)}' on line {line_num + 1}:\n"
+                                 f"Skipped malformed content: '{line.strip()}'\nReason: {e}\n")
+                    self.parsing_errors.append(error_msg)
+
 
             # Update progress after each image
             self.progress_updated.emit(i + 1)
@@ -290,11 +294,14 @@ class Base(QDialog):
         )
         if file_path:
             self.yaml_path_label.setText(file_path)
-            # Auto-fill output directory and folder name if not set
+            # Auto-fill output directory to be the PARENT of the yaml's directory
             if not self.output_dir_label.text():
-                self.output_dir_label.setText(os.path.dirname(file_path))
+                parent_dir = os.path.dirname(os.path.dirname(file_path))
+                self.output_dir_label.setText(parent_dir)
             if not self.output_folder_name.text():
-                self.output_folder_name.setText("project")
+                # Suggest a folder name based on the yaml file's parent folder
+                project_name = os.path.basename(os.path.dirname(file_path))
+                self.output_folder_name.setText(f"{project_name}_imported")
 
     def browse_output_dir(self):
         """Open a dialog to select the output directory."""
@@ -309,6 +316,16 @@ class Base(QDialog):
         if not all([self.yaml_path_label.text(), self.output_dir_label.text(), self.output_folder_name.text()]):
             QMessageBox.warning(self, "Error", "Please fill in all fields.")
             return
+        
+        # This check for existing output is still relevant
+        self.output_folder = os.path.join(self.output_dir_label.text(), self.output_folder_name.text())
+        if os.path.exists(self.output_folder) and os.listdir(self.output_folder):
+            reply = QMessageBox.question(self,
+                                         'Directory Not Empty',
+                                         f"The directory '{self.output_folder}' is not empty. Continue?",
+                                         QMessageBox.Yes | QMessageBox.No, QMessageBox.No)
+            if reply == QMessageBox.No:
+                return
 
         # Pre-scan for duplicates
         yaml_path = self.yaml_path_label.text()
@@ -358,16 +375,7 @@ class Base(QDialog):
                 rename_files = False
             else:  # User closed the dialog
                 return
-
-        self.output_folder = os.path.join(self.output_dir_label.text(), self.output_folder_name.text())
-        if os.path.exists(self.output_folder) and os.listdir(self.output_folder):
-            reply = QMessageBox.question(self,
-                                         'Directory Not Empty',
-                                         f"The directory '{self.output_folder}' is not empty. Continue?",
-                                         QMessageBox.Yes | QMessageBox.No, QMessageBox.No)
-            if reply == QMessageBox.No:
-                return
-
+        
         self.button_box.setEnabled(False)
         QApplication.setOverrideCursor(Qt.WaitCursor)
 
@@ -393,6 +401,7 @@ class Base(QDialog):
         self.worker.error.connect(self.on_error)
         self.worker.status_changed.connect(self.on_status_changed)
         self.worker.progress_updated.connect(self.on_progress_update)
+        # Connect to the updated signal
         self.worker.processing_complete.connect(self.on_processing_complete)
         self.thread.start()
 
@@ -403,7 +412,7 @@ class Base(QDialog):
     def on_progress_update(self, value):
         self.progress_bar.set_value(value)
 
-    def on_processing_complete(self, raw_annotations, image_paths):
+    def on_processing_complete(self, raw_annotations, image_paths, parsing_errors):
         added_paths = []
         for path in image_paths:
             if self.image_window.add_image(path):
@@ -447,9 +456,20 @@ class Base(QDialog):
             self.image_window.update_image_annotations(added_paths[-1])
             self.annotation_window.load_annotations()
 
-        QMessageBox.information(self, 
-                                "Dataset Imported",
-                                "Dataset has been successfully imported.")
+        # --- Display a summary message, including any parsing errors ---
+        summary_message = "Dataset has been successfully imported."
+        if parsing_errors:
+            # If there were errors, show a more detailed dialog
+            QMessageBox.warning(self, 
+                                "Import Complete with Warnings",
+                                f"{summary_message}\n\nHowever, {len(parsing_errors)} issue(s) were found "
+                                "in the label files. Please review them below.",
+                                details='\n'.join(parsing_errors))
+        else:
+            # Otherwise, show a simple info box
+            QMessageBox.information(self, 
+                                    "Dataset Imported",
+                                    summary_message)
 
     def export_annotations_to_json(self, annotations_list, output_dir):
         """

coralnet_toolbox/MachineLearning/VideoInference/QtBase.py

@@ -320,10 +320,6 @@ class Base(QDialog):
             # If video already loaded, update output dir for widget
             if self.video_path:
                 self.video_region_widget.load_video(self.video_path, dir_name)
-            else:
-                self.update_record_buttons()
-        else:
-            self.update_record_buttons()
 
     def browse_model(self):
         """Open file dialog to select model file (filtered to .pt, .pth)."""

coralnet_toolbox/QtAnnotationWindow.py

@@ -408,10 +408,6 @@ class AnnotationWindow(QGraphicsView):
 
         self.toggle_cursor_annotation()
 
-        # Set the image dimensions, and current view in status bar
-        self.imageLoaded.emit(self.pixmap_image.width(), self.pixmap_image.height())
-        self.viewChanged.emit(self.pixmap_image.width(), self.pixmap_image.height())
-
         # Load all associated annotations
         self.load_annotations()
         # Update the image window's image annotations
@@ -421,6 +417,10 @@ class AnnotationWindow(QGraphicsView):
 
         QApplication.processEvents()
 
+        # Set the image dimensions, and current view in status bar
+        self.imageLoaded.emit(self.pixmap_image.width(), self.pixmap_image.height())
+        self.viewChanged.emit(self.pixmap_image.width(), self.pixmap_image.height())
+
     def update_current_image_path(self, image_path):
         """Update the current image path being displayed."""
         self.current_image_path = image_path
@@ -466,29 +466,57 @@ class AnnotationWindow(QGraphicsView):
         self.centerOn(annotation_center)
     
     def center_on_annotation(self, annotation):
-        """Center and zoom in to focus on the specified annotation."""
+        """Center and zoom in to focus on the specified annotation with dynamic padding."""
         # Create graphics item if it doesn't exist
         if not annotation.graphics_item:
             annotation.create_graphics_item(self.scene)
 
         # Get the bounding rect of the annotation in scene coordinates
         annotation_rect = annotation.graphics_item.boundingRect()
-        
-        # Add some padding around the annotation (20% on each side)
-        padding_x = annotation_rect.width() * 0.2
-        padding_y = annotation_rect.height() * 0.2
+
+        # Step 1: Calculate annotation and image area
+        annotation_area = annotation_rect.width() * annotation_rect.height()
+        if self.pixmap_image:
+            image_width = self.pixmap_image.width()
+            image_height = self.pixmap_image.height()
+        else:
+            # Fallback to scene rect if image not loaded
+            image_width = self.scene.sceneRect().width()
+            image_height = self.scene.sceneRect().height()
+        image_area = image_width * image_height
+
+        # Step 2: Compute the relative area ratio (avoid division by zero)
+        if image_area > 0:
+            relative_area = annotation_area / image_area
+        else:
+            relative_area = 1.0  # fallback, treat as full image
+
+        # Step 3: Map ratio to padding factor (smaller annotation = more padding)
+        # Example: padding_factor = clamp(0.5 * (1/relative_area)**0.5, 0.1, 0.5)
+        # - For very small annotations, padding approaches 0.5 (50%)
+        # - For large annotations, padding approaches 0.1 (10%)
+        import math
+        min_padding = 0.1  # 10%
+        max_padding = 0.5  # 50%
+        if relative_area > 0:
+            padding_factor = max(min(0.5 * (1 / math.sqrt(relative_area)), max_padding), min_padding)
+        else:
+            padding_factor = min_padding
+
+        # Step 4: Apply dynamic padding
+        padding_x = annotation_rect.width() * padding_factor
+        padding_y = annotation_rect.height() * padding_factor
         padded_rect = annotation_rect.adjusted(-padding_x, -padding_y, padding_x, padding_y)
-        
+
         # Fit the padded annotation rect in the view
         self.fitInView(padded_rect, Qt.KeepAspectRatio)
-        
+
         # Update the zoom factor based on the new view transformation
-        # We can calculate this by comparing the viewport size to the scene rect size
         view_rect = self.viewport().rect()
-        zoom_x = view_rect.width() / padded_rect.width() 
+        zoom_x = view_rect.width() / padded_rect.width()
         zoom_y = view_rect.height() / padded_rect.height()
         self.zoom_factor = min(zoom_x, zoom_y)
-        
+
         # Signal that the view has changed
         self.viewChanged.emit(*self.get_image_dimensions())
 

coralnet_toolbox/QtEventFilter.py

@@ -23,6 +23,7 @@ class GlobalEventFilter(QObject):
         self.detect_deploy_model_dialog = main_window.detect_deploy_model_dialog
         self.segment_deploy_model_dialog = main_window.segment_deploy_model_dialog
         self.sam_deploy_generator_dialog = main_window.sam_deploy_generator_dialog
+        self.see_anything_deploy_generator_dialog = main_window.see_anything_deploy_generator_dialog
         self.auto_distill_deploy_model_dialog = main_window.auto_distill_deploy_model_dialog
         
     def eventFilter(self, obj, event):
@@ -69,9 +70,14 @@ class GlobalEventFilter(QObject):
                 if event.key() == Qt.Key_4:
                     self.sam_deploy_generator_dialog.predict()
                     return True
-
-                # Handle hotkey for auto distill prediction
+                
+                # Handle hotkey for see anything (YOLOE) generator
                 if event.key() == Qt.Key_5:
+                    self.see_anything_deploy_generator_dialog.predict()
+                    return True
+                
+                # Handle hotkey for auto distill prediction
+                if event.key() == Qt.Key_6:
                     self.auto_distill_deploy_model_dialog.predict()
                     return True
 
@@ -85,9 +91,20 @@ class GlobalEventFilter(QObject):
                 
             # Delete (backspace or delete key) selected annotations when select tool is active
             if event.key() == Qt.Key_Delete or event.key() == Qt.Key_Backspace:
+                # First check if the select tool is active
                 if self.main_window.select_tool_action.isChecked():
+                    selected_tool = self.annotation_window.selected_tool
+                    select_tool = self.annotation_window.tools[selected_tool]
+                    # Get the active subtool if it exists, pass to its keyPressEvent
+                    if hasattr(select_tool, 'active_subtool') and select_tool.active_subtool:
+                        select_tool.active_subtool.keyPressEvent(event)
+                        return True
+                            
+                    # Otherwise, proceed with deletion if there are selected annotations
                     if self.annotation_window.selected_annotations:
                         self.annotation_window.delete_selected_annotations()
+                        return True
+                    
                     # Consume the event so it doesn't do anything else
                     return True