""" Panel Extractor - Extracts and saves individual comic panels as 640x800 images """ import os import json import cv2 import numpy as np from PIL import Image, ImageDraw, ImageFont from typing import List, Dict, Tuple class PanelExtractor: def __init__(self, output_dir: str = "output/panels"): """Initialize panel extractor Args: output_dir: Directory to save extracted panels """ self.output_dir = output_dir self.panel_size = (640, 800) # Width x Height def extract_panels_from_comic(self, pages_json_path: str = "output/pages.json", frames_dir: str = "frames/final") -> List[str]: """Extract panels from generated comic data Args: pages_json_path: Path to pages.json file frames_dir: Directory containing frame images Returns: List of saved panel file paths """ # Create output directory os.makedirs(self.output_dir, exist_ok=True) # Clear existing panels for file in os.listdir(self.output_dir): if file.endswith('.jpg') or file.endswith('.png'): os.remove(os.path.join(self.output_dir, file)) # Load comic data try: with open(pages_json_path, 'r') as f: pages_data = json.load(f) except Exception as e: print(f"❌ Failed to load comic data: {e}") return [] saved_panels = [] panel_count = 0 print(f"📸 Extracting panels as {self.panel_size[0]}x{self.panel_size[1]} images...") # Process each page for page_idx, page in enumerate(pages_data): panels = page.get('panels', []) bubbles = page.get('bubbles', []) # Process each panel for panel_idx, panel in enumerate(panels): panel_count += 1 # Extract panel image panel_img = self._extract_panel(panel, frames_dir) if panel_img is None: continue # Find bubbles that belong to this panel panel_bubbles = self._find_panel_bubbles(panel, bubbles) # Add bubbles to panel if panel_bubbles: panel_img = self._add_bubbles_to_panel(panel_img, panel, panel_bubbles) # Resize to target size panel_img = self._resize_panel(panel_img) # Save panel filename = f"panel_{panel_count:03d}_p{page_idx+1}_{panel_idx+1}.jpg" filepath = os.path.join(self.output_dir, filename) # Convert to RGB if needed (remove alpha channel) if len(panel_img.shape) == 3 and panel_img.shape[2] == 4: panel_img = cv2.cvtColor(panel_img, cv2.COLOR_BGRA2BGR) elif len(panel_img.shape) == 2: panel_img = cv2.cvtColor(panel_img, cv2.COLOR_GRAY2BGR) cv2.imwrite(filepath, panel_img, [cv2.IMWRITE_JPEG_QUALITY, 95]) saved_panels.append(filepath) print(f"✅ Extracted {len(saved_panels)} panels to: {self.output_dir}") # Create an index HTML for viewing self._create_panel_viewer(saved_panels) return saved_panels def _extract_panel(self, panel: Dict, frames_dir: str) -> np.ndarray: """Extract panel region from frame image""" try: # Get frame path frame_filename = os.path.basename(panel['image']) frame_path = os.path.join(frames_dir, frame_filename) if not os.path.exists(frame_path): # Try without 'final' in path frame_path = panel['image'].lstrip('/') if not os.path.exists(frame_path): print(f"⚠️ Frame not found: {frame_path}") return None # Load frame frame = cv2.imread(frame_path) if frame is None: print(f"⚠️ Failed to load frame: {frame_path}") return None # No need to extract region - use full frame as is # The panel coordinates are for HTML display, not image cropping return frame except Exception as e: print(f"❌ Failed to extract panel: {e}") return None def _find_panel_bubbles(self, panel: Dict, bubbles: List[Dict]) -> List[Dict]: """Find speech bubbles that belong to a panel""" panel_bubbles = [] # Panel boundaries px1 = panel['x'] py1 = panel['y'] px2 = px1 + panel['width'] py2 = py1 + panel['height'] for bubble in bubbles: # Bubble center bx = bubble['x'] + bubble['width'] / 2 by = bubble['y'] + bubble['height'] / 2 # Check if bubble center is within panel if px1 <= bx <= px2 and py1 <= by <= py2: # Adjust bubble coordinates relative to panel adjusted_bubble = bubble.copy() adjusted_bubble['x'] -= px1 adjusted_bubble['y'] -= py1 panel_bubbles.append(adjusted_bubble) return panel_bubbles def _add_bubbles_to_panel(self, panel_img: np.ndarray, panel: Dict, bubbles: List[Dict]) -> np.ndarray: """Add speech bubbles to panel image""" # Convert to PIL for easier drawing img = Image.fromarray(cv2.cvtColor(panel_img, cv2.COLOR_BGR2RGB)) draw = ImageDraw.Draw(img) # Try to load a comic font try: font = ImageFont.truetype("/usr/share/fonts/truetype/liberation/LiberationSans-Bold.ttf", 16) except: font = None for bubble in bubbles: # Scale bubble coordinates to match image size img_h, img_w = panel_img.shape[:2] panel_w, panel_h = panel['width'], panel['height'] # Scale factors scale_x = img_w / panel_w scale_y = img_h / panel_h # Scaled coordinates x = int(bubble['x'] * scale_x) y = int(bubble['y'] * scale_y) w = int(bubble['width'] * scale_x) h = int(bubble['height'] * scale_y) # Draw bubble background bubble_bbox = [x, y, x + w, y + h] draw.ellipse(bubble_bbox, fill='white', outline='black', width=2) # Draw text text = bubble.get('text', '') if text and font: # Word wrap text words = text.split() lines = [] current_line = [] for word in words: current_line.append(word) line_text = ' '.join(current_line) bbox = draw.textbbox((0, 0), line_text, font=font) if bbox[2] > w - 20: # Leave padding if len(current_line) > 1: current_line.pop() lines.append(' '.join(current_line)) current_line = [word] else: lines.append(line_text) current_line = [] if current_line: lines.append(' '.join(current_line)) # Draw centered text line_height = 20 total_height = len(lines) * line_height start_y = y + (h - total_height) // 2 for i, line in enumerate(lines): bbox = draw.textbbox((0, 0), line, font=font) text_width = bbox[2] - bbox[0] text_x = x + (w - text_width) // 2 text_y = start_y + i * line_height draw.text((text_x, text_y), line, fill='black', font=font) # Convert back to numpy return cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR) def _resize_panel(self, panel_img: np.ndarray) -> np.ndarray: """Resize panel to target size (640x800)""" h, w = panel_img.shape[:2] target_w, target_h = self.panel_size # Calculate scale to fit within target size while maintaining aspect ratio scale = min(target_w / w, target_h / h) new_w = int(w * scale) new_h = int(h * scale) # Resize image resized = cv2.resize(panel_img, (new_w, new_h), interpolation=cv2.INTER_LANCZOS4) # Create canvas of target size canvas = np.ones((target_h, target_w, 3), dtype=np.uint8) * 255 # White background # Center the resized image on canvas x_offset = (target_w - new_w) // 2 y_offset = (target_h - new_h) // 2 canvas[y_offset:y_offset+new_h, x_offset:x_offset+new_w] = resized return canvas def _create_panel_viewer(self, panel_files: List[str]): """Create an HTML viewer for extracted panels""" html = ''' Extracted Comic Panels - 640x800

📸 Extracted Comic Panels (640x800)

All panels have been extracted and resized to 640x800 pixels

''' for panel_path in panel_files: filename = os.path.basename(panel_path) panel_num = filename.split('_')[1] html += f'''
{filename}
Panel {panel_num}
''' html += '''
''' viewer_path = os.path.join(self.output_dir, 'panel_viewer.html') with open(viewer_path, 'w', encoding='utf-8') as f: f.write(html) print(f"📄 Panel viewer created: {viewer_path}") # Convenience function for command line usage def extract_panels(pages_json: str = "output/pages.json", frames_dir: str = "frames/final", output_dir: str = "output/panels"): """Extract panels from comic""" extractor = PanelExtractor(output_dir) return extractor.extract_panels_from_comic(pages_json, frames_dir) if __name__ == "__main__": extract_panels()