1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
|
import os
from os.path import join
from pathlib import Path
import math
import cv2 as cv
import numpy as np
import imutils
from app.utils import im_utils, logger_utils
from app.models.bbox import BBox
from app.settings import app_cfg as cfg
from app.settings import types
class FacePoseDLIB:
dnn_size = (400, 400)
pose_types = {'pitch': (0,0,255), 'roll': (255,0,0), 'yaw': (0,255,0)}
def __init__(self):
self.log = logger_utils.Logger.getLogger()
def pose(self, landmarks_norm, dim):
'''Returns face pose information
:param landmarks: (list) of 68 (int, int) xy tuples
:param dim: (tuple|list) of image (width, height)
:returns (dict) of pose attributes
'''
# computes pose using 6 / 68 points from dlib face landmarks
# based on learnopencv.com and
# https://github.com/jerryhouuu/Face-Yaw-Roll-Pitch-from-Pose-Estimation-using-OpenCV/
# NB: not as accurate as MTCNN, see @jerryhouuu for ideas
pose_points_idx = (30, 8, 36, 45, 48, 54)
axis = np.float32([[500,0,0], [0,500,0], [0,0,500]])
# 3D model points.
model_points = np.array([
(0.0, 0.0, 0.0), # Nose tip
(0.0, -330.0, -65.0), # Chin
(-225.0, 170.0, -135.0), # Left eye left corner
(225.0, 170.0, -135.0), # Right eye right corne
(-150.0, -150.0, -125.0), # Left Mouth corner
(150.0, -150.0, -125.0) # Right mouth corner
])
# Assuming no lens distortion
dist_coeffs = np.zeros((4,1))
# find 6 pose points
pose_points = []
for j, idx in enumerate(pose_points_idx):
x,y = landmarks_norm[idx]
pt = (int(x*dim[0]), int(y*dim[1]))
pose_points.append(pt)
pose_points = np.array(pose_points, dtype='double') # convert to double, real dimensions
# create camera matrix
focal_length = dim[0]
center = (dim[0]/2, dim[1]/2)
cam_mat = np.array(
[[focal_length, 0, center[0]],
[0, focal_length, center[1]],
[0, 1, 1]], dtype = "double")
# solve PnP for rotation and translation
(success, rot_vec, tran_vec) = cv.solvePnP(model_points, pose_points,
cam_mat, dist_coeffs,
flags=cv.SOLVEPNP_ITERATIVE)
result = {}
# project points
pts_im, jac = cv.projectPoints(axis, rot_vec, tran_vec, cam_mat, dist_coeffs)
pts_model, jac2 = cv.projectPoints(model_points, rot_vec, tran_vec, cam_mat, dist_coeffs)
result['points'] = {
'pitch': list(map(int,pts_im[0][0])),
'roll': list(map(int,pts_im[2][0])),
'yaw': list(map(int,pts_im[1][0]))
}
result['point_nose'] = tuple(map(int,pose_points[0]))
rvec_matrix = cv.Rodrigues(rot_vec)[0]
# convert to degrees
proj_matrix = np.hstack((rvec_matrix, tran_vec))
eulerAngles = cv.decomposeProjectionMatrix(proj_matrix)[6]
pitch, yaw, roll = [math.radians(x) for x in eulerAngles]
pitch = math.degrees(math.asin(math.sin(pitch)))
roll = -math.degrees(math.asin(math.sin(roll)))
yaw = math.degrees(math.asin(math.sin(yaw)))
result['pitch'] = pitch
result['roll'] = roll
result['yaw'] = yaw
return result
|
