38 lines
1.8 KiB
Python
38 lines
1.8 KiB
Python
import numpy as np
|
|
import cv2
|
|
import mediapipe as mp
|
|
mp_face_mesh = mp.solutions.face_mesh
|
|
mp_face_detection = mp.solutions.face_detection
|
|
|
|
def detect_face_mesh(frames):
|
|
pts_3d = np.zeros([len(frames), 478, 3])
|
|
with mp_face_mesh.FaceMesh(
|
|
static_image_mode=True,
|
|
max_num_faces=1,
|
|
refine_landmarks=True,
|
|
min_detection_confidence=0.5) as face_mesh:
|
|
for frame_index, frame in enumerate(frames):
|
|
results = face_mesh.process(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
|
|
if results.multi_face_landmarks:
|
|
image_height, image_width = frame.shape[:2]
|
|
for face_landmarks in results.multi_face_landmarks:
|
|
for index_, i in enumerate(face_landmarks.landmark):
|
|
x_px = i.x * image_width
|
|
y_px = i.y * image_height
|
|
z_px = i.z * image_width
|
|
pts_3d[frame_index, index_] = np.array([x_px, y_px, z_px])
|
|
else:
|
|
break
|
|
return pts_3d
|
|
def detect_face(frames):
|
|
rect_2d = np.zeros([len(frames), 4])
|
|
# 剔除掉多个人脸、大角度侧脸（鼻子不在两个眼之间）、部分人脸框在画面外、人脸像素低于80*80的
|
|
with mp_face_detection.FaceDetection(
|
|
model_selection=1, min_detection_confidence=0.5) as face_detection:
|
|
for frame_index, frame in enumerate(frames):
|
|
results = face_detection.process(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
|
|
if not results.detections or len(results.detections) > 1:
|
|
break
|
|
rect = results.detections[0].location_data.relative_bounding_box
|
|
rect_2d[frame_index] = np.array([rect.xmin, rect.xmin + rect.width, rect.ymin, rect.ymin + rect.height])
|
|
return rect_2d |