File size: 11,260 Bytes
2c68f56 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 | import cv2
import numpy as np
import mediapipe as mp
from typing import Mapping
try:
from PIL.Image import Resampling
RESAMPLING_METHOD = Resampling.BICUBIC
except ImportError:
from PIL.Image import BICUBIC
RESAMPLING_METHOD = BICUBIC
import torch
def number_to_rgb(n):
n = int(n)
r = (n >> 16) & 0xFF
g = (n >> 8) & 0xFF
b = n & 0xFF
return (r, g, b)
# cv2.imwrite(os.path.join(save_path,f"smplx_stp_{iter_idx}_{im_idx}.png"), cv2.cvtColor(_img.copy(), cv2.COLOR_RGB2BGR))
def draw_landmarks(landmarks, image, color=(0, 255, 0), radius=4, thickness=-1, viz_index=False):
ret_img = image.copy()
for (idx, lmk) in enumerate(landmarks):
if len(lmk) > 2:
color = number_to_rgb(lmk[2]*1500)
ret_img = cv2.circle(ret_img, (int(lmk[0]), int(lmk[1])), radius, color, thickness) # 像素点坐标从 0-1024, x 为横坐标 w,y 为h坐标
if viz_index:
cv2.putText(ret_img, f'{idx}', (int(lmk[0]), int(lmk[1])), cv2.FONT_HERSHEY_COMPLEX_SMALL, 2, ((idx*2)%255, 145, (idx*3)%255), 1)
return ret_img
def _draw_mp_kps(image, landmarks, connections, landmark_drawing_spec, connection_drawing_spec):
t_landmarks = np.int32(landmarks[:, :2])
for connection in connections:
start_idx = connection[0]
end_idx = connection[1]
drawing_spec = connection_drawing_spec[connection] if isinstance(
connection_drawing_spec, Mapping) else connection_drawing_spec
cv2.line(image, t_landmarks[start_idx],
t_landmarks[end_idx], drawing_spec.color,
drawing_spec.thickness)
if landmark_drawing_spec:
for idx in range(t_landmarks.shape[0]):
landmark_px = t_landmarks[idx]
drawing_spec = landmark_drawing_spec[idx] if isinstance(
landmark_drawing_spec, Mapping) else landmark_drawing_spec
# White circle border
circle_border_radius = max(drawing_spec.circle_radius + 1,
int(drawing_spec.circle_radius * 1.2))
cv2.circle(image, landmark_px, circle_border_radius, (224, 224, 224),
drawing_spec.thickness)
# Fill color into the circle
cv2.circle(image, landmark_px, drawing_spec.circle_radius,
drawing_spec.color, drawing_spec.thickness)
mp_drawing = mp.solutions.drawing_utils
mp_drawing_styles = mp.solutions.drawing_styles
mp_connections = mp.solutions.face_mesh_connections
def draw_mediapipe_kps(mp_kps, canvas):
ret_img = canvas.copy()
# draw FACEMESH_TESSELATION
_draw_mp_kps(ret_img, mp_kps,
mp.solutions.face_mesh.FACEMESH_TESSELATION,
connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style(),
landmark_drawing_spec=None)
# draw FACEMESH_CONTOURS
_draw_mp_kps(ret_img, mp_kps,
mp.solutions.face_mesh.FACEMESH_CONTOURS,
connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_contours_style(),
landmark_drawing_spec=None)
# draw FACEMESH_IRISES
_draw_mp_kps(ret_img, mp_kps,
mp.solutions.face_mesh.FACEMESH_IRISES,
connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_iris_connections_style(),
landmark_drawing_spec=None)
return ret_img
def get_semantic_indices():
semantic_connections = {
'Contours': mp_connections.FACEMESH_CONTOURS,
'FaceOval': mp_connections.FACEMESH_FACE_OVAL,
'LeftIris': mp_connections.FACEMESH_LEFT_IRIS,
'LeftEye': mp_connections.FACEMESH_LEFT_EYE,
'LeftEyebrow': mp_connections.FACEMESH_LEFT_EYEBROW,
'RightIris': mp_connections.FACEMESH_RIGHT_IRIS,
'RightEye': mp_connections.FACEMESH_RIGHT_EYE,
'RightEyebrow': mp_connections.FACEMESH_RIGHT_EYEBROW,
'Lips': mp_connections.FACEMESH_LIPS,
'Tesselation': mp_connections.FACEMESH_TESSELATION
}
def get_compact_idx(connections):
ret = []
for conn in connections:
ret.append(conn[0])
ret.append(conn[1])
return sorted(tuple(set(ret)))
semantic_indexes = {k: get_compact_idx(v) for k, v in semantic_connections.items()}
return semantic_indexes
mp_lip_indices = [61, 185, 40, 39, 37, 0, 267, 269, 270, 409, 291, 375, 321, 405, 314, 17, 84, 181, 91, 146,
76, 184, 74, 73, 72, 11, 302, 303, 304, 408, 306, 307, 320, 404, 315, 16, 85, 180, 90, 77,
62, 183, 42, 41, 38, 12, 268, 271, 272, 407, 292, 325, 319, 403, 316, 15, 86, 179, 89, 96,
78, 191, 80, 81, 82, 13, 312, 311, 310, 415, 308, 324, 318, 402, 317, 14, 87, 178, 88, 95]
mp_lowerface_indices = [177, 132, 215, 58, 172, 136, 150, 149, 176, 148, 152, 377, 400, 378, 379, 365, 397, 288, 435, 361, 401,
147, 213, 192, 138, 135, 169, 170, 140, 171, 175, 396, 369, 395, 394, 364, 367, 433, 416, 376,
57, 186, 92, 165, 167, 164, 393, 391, 322, 410, 287, 273, 335, 406, 313, 18, 83, 182, 106, 43,
212, 216, 206, 203, 423, 426, 436, 432, 422, 424, 418, 421, 200, 201, 194, 204, 202,
214, 207, 205, 425, 427, 434, 430, 431, 262, 428, 199, 208, 32, 211, 210] # todo: need further check indices
chain_indices = [93, 137, 123, 50, 205, 206, 165, 167, 164, 393, 391, 426, 425, 280, 352, 366, 323]
teeth_indices = [62, 183, 42, 41, 38, 12, 268, 271, 272, 407, 292, 325, 319, 403, 316, 15, 86, 179, 89, 96]
nose_indices = [142, 129, 98, 97, 2, 326, 327, 358, 371, 355, 437, 399, 419, 197, 196, 174, 217, 126,
102, 64, 240, 99, 60, 75, 59, 235, 166, 219, 48, 49, 209,
331, 294, 460, 328, 290, 305, 289, 392, 439, 278, 279, 360,
198, 131, 115, 218, 79, 20, 242, 141, 94, 370, 462, 250, 309, 438, 344, 360, 420,
236, 3, 195, 248, 456, 363, 440, 457, 459, 458, 461, 354, 19, 125, 241, 238, 239, 237, 220, 134,
51, 5, 281, 275, 274, 354, 19, 125, 44, 45, 1, 4]
kp68_lowerface_indices = [x for x in range(3, 14)] + [52, 50, 3]
kp68_mouth_indices = [x for x in range(48, 61)]
def _merge_with_weight(p1, p2, w):
return p1 * (1 - w) + p2 * w
def merge_lower_face_pints(pred_lst, pred_ff_lst, ff_only=False):
if ff_only is True:
w1, w2 = 0.95, 0.95
else:
w1, w2 = 0.8, 0.2
pred_lst[mp_lip_indices] = _merge_with_weight(pred_lst[mp_lip_indices], pred_ff_lst[:len(mp_lip_indices)], w1)
pred_lst[mp_lowerface_indices] = _merge_with_weight(pred_lst[mp_lowerface_indices], pred_ff_lst[len(mp_lip_indices):], w2)
return pred_lst
def mask_from_points(size, points, radius=0, is_converx=True, mean_y=-1, ratio=0.025):
""" Create a mask of supplied size from supplied points
:param size: tuple of output mask size
:param points: array of [x, y] points
:returns: mask of values 0 and 255 where
255 indicates the convex hull containing the points
"""
if radius == 0:
radius = int(size[0] * ratio)
kernel = np.ones((int(abs(radius)), int(abs(radius))), np.uint8)
if mean_y > 0:
v_ids = points[:, :, 1] > mean_y
points[v_ids, 1] = mean_y + (points[v_ids, 1] - mean_y) * 1.3
mask = np.zeros(size, np.uint8)
if is_converx:
p = cv2.convexHull(points)
cv2.fillConvexPoly(mask, np.squeeze(p).astype(np.int32), 255)
else:
cv2.fillPoly(mask, [points.astype(np.int32)], 255)
if radius < 0:
mask = cv2.erode(mask, kernel)
else:
mask = cv2.dilate(mask, kernel)
return mask
def draw_teeth_mask(ver, canvas):
teeth_mask = mask_from_points(canvas.shape[:2], ver[:, teeth_indices, ...], radius=3)
teeth_mask = cv2.merge([np.zeros_like(teeth_mask), teeth_mask, teeth_mask])
canvas = np.clip(canvas / 1.0 + teeth_mask / 2.0, 0, 225)
return canvas.astype(np.uint8)
def draw_nose_mask(size, ver):
nose_mask = mask_from_points(size, ver[:, nose_indices, ...])
return nose_mask
def draw_lowerface_mask(size, ver, mean_y, is_kp68=False):
if not is_kp68:
lf_idx = mp_lowerface_indices + mp_lip_indices + chain_indices
lf_mask = mask_from_points(size, ver[:, lf_idx, ...], mean_y=mean_y)
else:
lf_idx = kp68_lowerface_indices
lf_mask = mask_from_points(size, ver[:, lf_idx, ...], is_converx=False, mean_y=mean_y)
return lf_mask
def draw_mouth_mask(size, ver, is_kp68=False):
if not is_kp68:
lf_idx = mp_lip_indices
lf_mask = mask_from_points(size, ver[:, lf_idx, ...])
else:
lf_idx = kp68_mouth_indices
lf_mask = mask_from_points(size, ver[:, lf_idx, ...], radius=27, is_converx=False)
return lf_mask
def draw_flame_lowerface_condition(img_fg:np.ndarray, img_fg_alpha:np.ndarray, img_bg:np.ndarray, lmk203:np.ndarray):
"""draw flame lower face condition
Args:
img_fg (np.ndarray): foreground image in 0~1, HxWxC
img_fg_alpha (np.ndarray): foreground image alpha in 0~1, HxW
img_bg (np.ndarray): background image in 0~1, HxWxC
lmk203 (np.ndarray): landmark in 203x2
"""
## 1. draw lower mask
lower_pts = np.concatenate([lmk203[114: 139], ((lmk203[57] + lmk203[202])/2)[None, ...]], axis=0)
# # enlarge lower y pts
lower_pts[:, 1] = (lower_pts[:, 1] - lmk203[201, 1]) * 1.2 + lmk203[201, 1]
mask1 = mask_from_points(img_bg.shape[:2], lower_pts, ratio=0.03, is_converx=False) / 255 # todo: check this and regenerate condition images
mouth_region = mask_from_points(img_bg.shape[:2], lmk203[84: 108], ratio=0.015, is_converx=False) / 255
mask1[mouth_region>0.05] = img_fg_alpha[mouth_region>0.05]
final_mask = np.clip(img_fg_alpha + mask1, 0, 1)
final_mask = cv2.merge([final_mask]*3)
## 2. draw inner mouth region
bg_img = img_bg * 255
bg_img[mouth_region>0.05] = (225, 225, 0)
## 3. overlay them
ret_img = final_mask * img_fg + (1 - final_mask) * bg_img / 255
return np.clip(ret_img, 0, 1)
def draw_fullface_mask(size, ver, mean_y):
lf_idx = mp_lowerface_indices + mp_lip_indices + chain_indices
lf_mask = mask_from_points(size, ver[:, lf_idx, ...], mean_y=mean_y)
ff_mask = mask_from_points(size, ver)
ff_mask = np.uint8(((lf_mask / 1.0 + ff_mask / 1.0) > 0) * 255.)
return ff_mask
def get_bbox_from_vert(vert):
rect = [vert[:,0].min(), vert[:,1].min(), vert[:,0].max(), vert[:,1].max()]
cx = rect[0] + (rect[2] - rect[0]) / 2
cy = rect[1] + (rect[3] - rect[1]) / 2
size = max(rect[2]- rect[0], rect[3]- rect[1])
size = size * 1.3
return [int(cx - size/2), int(cy-size/2), int(cx + size/2), int(cy + size/2)]
def alpha_feathering(src_img, dest_img, img_mask, use_blur=True):
if use_blur:
blur_radius = int(src_img.shape[0]/30)
mask = cv2.blur(img_mask, (blur_radius, blur_radius))
else:
mask = img_mask
mask = mask / 255.0
result_img = np.empty(src_img.shape, np.uint8)
for i in range(3):
result_img[..., i] = src_img[..., i] * mask + dest_img[..., i] * (1-mask)
return result_img
|