本文以人脸识别问题中的关键点检测问题为例，讲述不同的数据增强方式对结果的影响
文章目录- 一、模拟强光照射到人脸上效果
- 二、模拟人脸的阴影干扰
- 三、模拟掩膜对人脸图像的遮挡效果
- 结果
提示：以下是本篇文章正文内容，下面案例可供参考
一、模拟强光照射到人脸上效果示例：现实生活中，人脸收到光照的强度，对受到光照人脸的角度和人脸面部的起伏的影响。在本文中模拟了人收到电光源的干扰，如类似手电筒直接照到人脸区域。我们主要需要确定两个参数，一个是形成的光源分布的有效区域的半径，一个是受到光源照射的中心点坐标。然后通过计算在半径内点像素到中心像素的距离，通过对应的距离赋予不同权重，越靠近点光源中心，赋值强度最高。
def En(image):
x, y,_ = image.shape # 获取图片大小
radius = np.random.randint(10, int(min(x, y)), 1) #
pos_x = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心点坐标
pos_y = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心坐标
pos_x = int(pos_x[0])
pos_y = int(pos_y[0])
radius = int(radius[0])
strength = 100
for j in range(pos_y - radius, pos_y + radius):
for i in range(pos_x-radius, pos_x+radius):
distance = math.pow((pos_x - i), 2) + math.pow((pos_y - j), 2)
distance = np.sqrt(distance)
if distance < radius:
result = 1 - distance / radius
result = result*strength
# print(result)
image[i, j, 0] = min((image[i, j, 0] + result),255)
image[i, j, 1] = min((image[i, j, 1] + result),255)
image[i, j, 2] = min((image[i, j, 2] + result),255)
image = image.astype(np.uint8)
return image
二、模拟人脸的阴影干扰
显示中，当人脸区域出现阴影，说明人脸区域前方有物体遮挡了部分光源，使得被遮挡部分光源相对于其他部分显的更暗一些。而且正常情况下越垂直于和靠近于遮挡物的中心，人脸受到的阴影干扰越强。在本文中，我们将人脸光照散布的赋值方式转换成了减值的方式。
代码如下（示例）：
def De(image):
x, y,_ = image.shape # 获取图片大小
radius = np.random.randint(10, int(min(x, y)), 1) #
pos_x = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心点坐标
pos_y = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心坐标
pos_x = int(pos_x[0])
pos_y = int(pos_y[0])
radius = int(radius[0])
strength = 100
for j in range(pos_y - radius, pos_y + radius):
for i in range(pos_x-radius, pos_x+radius):
distance = math.pow((pos_x - i), 2) + math.pow((pos_y - j), 2)
distance = np.sqrt(distance)
if distance < radius:
result = 1 - distance / radius
result = result*strength
# print(result)
image[i, j, 0] = max((image[i, j, 0] - result),0)
image[i, j, 1] = max((image[i, j, 1] - result),0)
image[i, j, 2] = max((image[i, j, 2] - result),0)
image = image.astype(np.uint8)
return image
三、模拟掩膜对人脸图像的遮挡效果
def Ma(image):
x, y,_ = image.shape # 获取图片大小
mask_size = np.random.randint(10, 50, 1)
pos_x = np.random.randint(10, (min(x, y) - 50), 1) # 获取人脸光照区域的中心点坐标
pos_y = np.random.randint(10, (min(x, y) - 50), 1) # 获取人脸光照区域的中心坐标
pos_x = int(pos_x[0])
pos_y = int(pos_y[0])
mask_size = int(mask_size[0])
image[pos_x:pos_x + mask_size, pos_y:pos_y + mask_size] = 0
return image
完整代码：
import cv2 as cv
import math
import numpy as np
from copy import deepcopy
#读取原始图像
def En(image):
x, y,_ = image.shape # 获取图片大小
radius = np.random.randint(10, int(min(x, y)), 1) #
pos_x = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心点坐标
pos_y = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心坐标
pos_x = int(pos_x[0])
pos_y = int(pos_y[0])
radius = int(radius[0])
strength = 100
for j in range(pos_y - radius, pos_y + radius):
for i in range(pos_x-radius, pos_x+radius):
distance = math.pow((pos_x - i), 2) + math.pow((pos_y - j), 2)
distance = np.sqrt(distance)
if distance < radius:
result = 1 - distance / radius
result = result*strength
# print(result)
image[i, j, 0] = min((image[i, j, 0] + result),255)
image[i, j, 1] = min((image[i, j, 1] + result),255)
image[i, j, 2] = min((image[i, j, 2] + result),255)
image = image.astype(np.uint8)
return image
def De(image):
x, y,_ = image.shape # 获取图片大小
radius = np.random.randint(10, int(min(x, y)), 1) #
pos_x = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心点坐标
pos_y = np.random.randint(0, (min(x, y) - radius), 1) # 获取人脸光照区域的中心坐标
pos_x = int(pos_x[0])
pos_y = int(pos_y[0])
radius = int(radius[0])
strength = 100
for j in range(pos_y - radius, pos_y + radius):
for i in range(pos_x-radius, pos_x+radius):
distance = math.pow((pos_x - i), 2) + math.pow((pos_y - j), 2)
distance = np.sqrt(distance)
if distance < radius:
result = 1 - distance / radius
result = result*strength
# print(result)
image[i, j, 0] = max((image[i, j, 0] - result),0)
image[i, j, 1] = max((image[i, j, 1] - result),0)
image[i, j, 2] = max((image[i, j, 2] - result),0)
image = image.astype(np.uint8)
return image
def Ma(image):
x, y,_ = image.shape # 获取图片大小
mask_size = np.random.randint(10, 50, 1)
pos_x = np.random.randint(10, (min(x, y) - 50), 1) # 获取人脸光照区域的中心点坐标
pos_y = np.random.randint(10, (min(x, y) - 50), 1) # 获取人脸光照区域的中心坐标
pos_x = int(pos_x[0])
pos_y = int(pos_y[0])
mask_size = int(mask_size[0])
image[pos_x:pos_x + mask_size, pos_y:pos_y + mask_size] = 0
return image
img = cv.imread('4.png')
img1 = deepcopy(img)
img1 = En(img1)
img2 = deepcopy(img)
img2 = De(img2)
img3 = deepcopy(img)
img3 = Ma(img3)
Light_En = np.hstack([img,img1,img2,img3])
cv.imshow('face_enhanced', Light_En)
cv.waitKey(0)
cv.destroyAllWindows()
结果
在每一张图像的增强过程中，我对每个增强操作的位置和增强的范围大小进行了随机化，这样使得每次增强后的结果不一样，达到数据增强的效果。喜欢就点个赞呗，best wishes！
