import cv2 import numpy as np import os from tqdm import tqdm import re import concurrent.futures def get_video_duration(stderr_output): """Get the duration of a video using ffmpeg.""" ffmpeg_output = stderr_output match = re.search(r"Duration: (\d{2}):(\d{2}):(\d{2})\.(\d{2})", ffmpeg_output, re.IGNORECASE) if match: hours = int(match.group(1)) minutes = int(match.group(2)) seconds = int(match.group(3)) milliseconds = int(match.group(4)) total_seconds = hours * 3600 + minutes * 60 + seconds + milliseconds / 100 return total_seconds else: print("Duration not found in ffmpeg output.") return 0 def get_video_dimensions(stderr_output): """Extract video width and height from ffmpeg stderr output.""" match = re.search(r'(\d{3,4})x(\d{3,4})', stderr_output) if match: width = int(match.group(1)) height = int(match.group(2)) return width, height return None, None def get_video_fps(stderr_output): """Extract video frame rate from ffmpeg stderr output.""" match = re.search(r'(\d+(\.\d+)?) fps', stderr_output) if match: fps = float(match.group(1)) return fps return None def calculate_mse(frame1, frame2): """Calculate Mean Squared Error between two frames.""" frame1 = cv2.resize(frame1, (frame2.shape[1], frame2.shape[0])) # Ensure consistent dimensions mse = np.sum((frame1 - frame2) ** 2) / float(frame1.size) # Compute MSE return mse def is_static(mse, threshold=5): # Using slightly lower threshold """Determine if frame is static based on MSE threshold.""" return mse < threshold # Frame considered static if MSE below threshold def load_frame(frame_path): """Load a frame and convert to grayscale.""" frame = cv2.imread(frame_path) if frame is None: return None # Return None if frame loading fails return cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) def detect_static_video(frame_path, frame_count=20): """Detect if video is static by analyzing frame differences.""" try: static_frame_count = 0 sample_count = 0 # Generate frame file paths frame_paths = [os.path.join(frame_path, f'{frame_index*4:06d}.jpg') for frame_index in range(1, frame_count + 1)] # Parallel frame loading using multithreading with concurrent.futures.ThreadPoolExecutor() as executor: frames = list(executor.map(load_frame, frame_paths)) # Filter out failed frame loads frames = [frame for frame in frames if frame is not None] # Require minimum 2 frames for analysis if len(frames) < 2: return False # Insufficient frames for static detection # Count static frames prev_frame = frames[0] for gray_frame in frames[1:]: mse = calculate_mse(prev_frame, gray_frame) if is_static(mse): static_frame_count += 1 prev_frame = gray_frame # Update previous frame # Calculate static frame ratio sample_count = len(frames) # Actual processed frame count static_ratio = static_frame_count / sample_count if sample_count > 0 else 0 # Video considered static if >85% frames are static return static_ratio > 0.85 except Exception as e: print(f"Error processing video: {frame_path}, Error: {e}") return None # Return None to indicate error def process_video_folder(folder_path, output_file, error_log): """Batch process video folders for static detection.""" video_files = [f for f in os.listdir(folder_path) if re.search(r'^[a-zA-Z0-9_-]*_.\d*0$', f)] with open(output_file, 'w') as output: for idx, video_file in enumerate(tqdm(video_files, desc="Processing Videos", unit="file")): frame_path = os.path.join(folder_path, video_file) result = detect_static_video(frame_path) if result is None: # Log failed processing attempts with open(error_log, 'a') as error_output: error_output.write(f"Error processing: {video_file}\n") elif result: output.write(f"{video_file}\n") # Record static video filename print(f"{video_file} is static") # Example usage if __name__ == "__main__": folder_path = '' # Video frame directory output_file = './static_new.txt' # Static video output list error_log = './static_new_err.txt' # Error log file process_video_folder(folder_path, output_file, error_log)