最后编辑于2022-04
AI Studio 项目地址 : https://aistudio.baidu.com/aistudio/projectdetail/402824
效果
代码
以下代码是在 notebook 中的代码,如果是在本地执行,推荐使用 opencv 进行图像展示
首先需要安装 paddlehub, -q 表示静默安装
!pip install -q paddlehub==1.6.1
然后简单测试一下 paddlehub 的关键点检测情况
import cv2
import paddlehub as hub
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np
import math
%matplotlib inline
src_img = cv2.imread('face/01.jpg')
module = hub.Module(name="face_landmark_localization")
result = module.keypoint_detection(images=[src_img])
tmp_img = src_img.copy()
for index, point in enumerate(result[0]['data'][0]):
# print(point)
# cv2.putText(img, str(index), (int(point[0]), int(point[1])), cv2.FONT_HERSHEY_COMPLEX, 3, (0,0,255), -1)
cv2.circle(tmp_img, (int(point[0]), int(point[1])), 2, (0, 0, 255), -1)
res_img_path = 'face_landmark.jpg'
cv2.imwrite(res_img_path, tmp_img)
img = mpimg.imread(res_img_path)
# 展示预测68个关键点结果
plt.figure(figsize=(10,10))
plt.imshow(img)
plt.axis('off')
plt.show()
下面的代码是融合两张图像的函数
def overlay_transparent(background_img, img_to_overlay_t, x, y, overlay_size=None):
bg_img = background_img.copy()
# convert 3 channels to 4 channels
if bg_img.shape[2] == 3:
bg_img = cv2.cvtColor(bg_img, cv2.COLOR_BGR2BGRA)
if overlay_size is not None:
img_to_overlay_t = cv2.resize(img_to_overlay_t.copy(), overlay_size)
b, g, r, a = cv2.split(img_to_overlay_t)
mask = cv2.medianBlur(a, 5)
h, w, _ = img_to_overlay_t.shape
roi = bg_img[int(y - h / 2):int(y + h / 2), int(x - w / 2):int(x + w / 2)]
img1_bg = cv2.bitwise_and(roi.copy(), roi.copy(), mask=cv2.bitwise_not(mask))
img2_fg = cv2.bitwise_and(img_to_overlay_t, img_to_overlay_t, mask=mask)
bg_img[int(y - h / 2):int(y + h / 2), int(x - w / 2):int(x + w / 2)] = cv2.add(img1_bg, img2_fg)
# convert 4 channels to 3 channels
bg_img = cv2.cvtColor(bg_img, cv2.COLOR_BGRA2BGR)
return bg_img
以下函数功能为计算两个点的角度,之后在旋转贴纸时会用到
from math import degrees, atan2
def angle_between(p1, p2):
x_diff = p2[0] - p1[0]
y_diff = p2[1] - p1[1]
return degrees(atan2(y_diff, x_diff))
以下函数通过将贴纸旋转,然后粘贴到对应的位置
def wear_glasses(image, glasses, eye_left_center, eye_right_center):
eye_left_center = np.array(eye_left_center)
eye_right_center = np.array(eye_right_center)
glasses_center = np.mean([eye_left_center, eye_right_center], axis=0) # put glasses's center to this center
glasses_size = np.linalg.norm(eye_left_center - eye_right_center) * 2 # the width of glasses mask
angle = -angle_between(eye_left_center, eye_right_center)
glasses_h, glasses_w = glasses.shape[:2]
glasses_c = (glasses_w / 2, glasses_h / 2)
M = cv2.getRotationMatrix2D(glasses_c, angle, 1)
cos = np.abs(M[0, 0])
sin = np.abs(M[0, 1])
# compute the new bounding dimensions of the image
nW = int((glasses_h * sin) + (glasses_w * cos))
nH = int((glasses_h * cos) + (glasses_w * sin))
# adjust the rotation matrix to take into account translation
M[0, 2] += (nW / 2) - glasses_c[0]
M[1, 2] += (nH / 2) - glasses_c[1]
rotated_glasses = cv2.warpAffine(glasses, M, (nW, nH))
try:
image = overlay_transparent(image, rotated_glasses, glasses_center[0], glasses_center[1],
overlay_size=(
int(glasses_size),
int(rotated_glasses.shape[0] * glasses_size / rotated_glasses.shape[1]))
)
except:
print('failed overlay image')
return image
以下函数用于计算眼睛的位置,用于计算出贴纸的尺寸和位置
def get_eye_center_point(landmarks, idx1, idx2):
center_x = (landmarks[idx1][0] + landmarks[idx2][0]) // 2
center_y = (landmarks[idx1][1] + landmarks[idx2][1]) // 2
return (center_x, center_y)
最后是主函数,在 notebook 中展示动态效果图
import os
import matplotlib.animation as animation
from IPython.display import HTML
glasses_lists = []
fig = plt.figure()
module = hub.Module(name="face_landmark_localization")
for path in os.listdir('glasses'):
image_file = 'face/01.jpg'
glasses_file = './glasses/' + path
image = cv2.imread(image_file)
glasses = cv2.imread(glasses_file, cv2.IMREAD_UNCHANGED)
result = module.keypoint_detection(images=[image])
landmarks = result[0]['data'][0]
eye_left_point = get_eye_center_point(landmarks, 36, 39)
eye_right_point = get_eye_center_point(landmarks, 42, 45)
image = wear_glasses(image, glasses, eye_left_point, eye_right_point)
image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
im = plt.imshow(image, animated=True)
plt.axis('off')
glasses_lists.append([im])
ani = animation.ArtistAnimation(fig, glasses_lists, interval=1000, blit=True, repeat_delay=1000)
HTML(ani.to_html5_video())
最终效果可以进入到 AI Studio 项目中进行查看
收藏
点赞
0
个赞
请登录后评论
TOP
切换版块
可以通过变量控制眼镜的样式,训练的时候学习对应眼镜的样式
不懂 GAN,没学过。用 GAN 直接生成想要的贴纸么
这个用gan做估计也可以