import face_detection import numpy as np import cv2 from tqdm import tqdm import torch import glob import os from natsort import natsorted device = "cuda" if torch.cuda.is_available() else "cpu" def get_squre_coords(coords, image, size=None, last_size=None): y1, y2, x1, x2 = coords w, h = x2 - x1, y2 - y1 center = (x1 + w // 2, y1 + h // 2) if size is None: size = (w + h) // 2 if last_size is not None: size = (w + h) // 2 size = (size - last_size) // 5 + last_size x1, y1 = center[0] - size // 2, center[1] - size // 2 x2, y2 = x1 + size, y1 + size return size, [y1, y2, x1, x2] def get_smoothened_boxes(boxes, T): for i in range(len(boxes)): if i + T > len(boxes): window = boxes[len(boxes) - T :] else: window = boxes[i : i + T] boxes[i] = np.mean(window, axis=0) return boxes def face_detect(images, pads): detector = face_detection.FaceAlignment(face_detection.LandmarksType._2D, flip_input=False, device=device) batch_size = 32 if device == "cuda" else 4 print("face detect batch size:", batch_size) while 1: predictions = [] try: for i in tqdm(range(0, len(images), batch_size)): predictions.extend(detector.get_detections_for_batch(np.array(images[i : i + batch_size]))) except RuntimeError: if batch_size == 1: raise RuntimeError("Image too big to run face detection on GPU. Please use the --resize_factor argument") batch_size //= 2 print("Recovering from OOM error; New batch size: {}".format(batch_size)) continue break results = [] pady1, pady2, padx1, padx2 = pads for rect, image in zip(predictions, images): if rect is None: cv2.imwrite(".temp/faulty_frame.jpg", image) # check this frame where the face was not detected. raise ValueError("Face not detected! Ensure the video contains a face in all the frames.") y1 = max(0, rect[1] - pady1) y2 = min(image.shape[0], rect[3] + pady2) x1 = max(0, rect[0] - padx1) x2 = min(image.shape[1], rect[2] + padx2) # y_gap, x_gap = ((y2 - y1) * 2) // 3, ((x2 - x1) * 2) // 3 y_gap, x_gap = (y2 - y1) // 2, (x2 - x1) // 2 coords_ = [y1 - y_gap, y2 + y_gap, x1 - x_gap, x2 + x_gap] _, coords = get_squre_coords(coords_, image) y1, y2, x1, x2 = coords y1 = max(0, y1) y2 = min(image.shape[0], y2) x1 = max(0, x1) x2 = min(image.shape[1], x2) results.append([x1, y1, x2, y2]) print("Number of frames cropped: {}".format(len(results))) print("First coords: {}".format(results[0])) boxes = np.array(results) boxes = get_smoothened_boxes(boxes, T=25) # results = [[image[y1:y2, x1:x2], (y1, y2, x1, x2)] for image, (x1, y1, x2, y2) in zip(images, boxes)] del detector return boxes def add_black(imgs): for i in range(len(imgs)): imgs[i] = cv2.vconcat([np.zeros((100, imgs[i].shape[1], 3), dtype=np.uint8), imgs[i], np.zeros((20, imgs[i].shape[1], 3), dtype=np.uint8)]) return imgs def preprocess(video_dir="./assets/videos", save_dir="./assets/coords"): all_videos = natsorted(glob.glob(os.path.join(video_dir, "*.mp4"))) for video_path in all_videos: video_stream = cv2.VideoCapture(video_path) # print('Reading video frames...') full_frames = [] while 1: still_reading, frame = video_stream.read() if not still_reading: video_stream.release() break full_frames.append(frame) print("Number of frames available for inference: " + str(len(full_frames))) full_frames = add_black(full_frames) # print('Face detection running...') coords = face_detect(full_frames, pads=(0, 0, 0, 0)) np.savez_compressed(os.path.join(save_dir, os.path.basename(video_path).split(".")[0]), coords=coords) def load_from_npz(video_name, save_dir="./assets/coords"): npz = np.load(os.path.join(save_dir, video_name + ".npz")) return npz["coords"] if __name__ == "__main__": preprocess()