Question

鉴于random picture，我已经成功检测到人脸并裁剪了计算出的ROI。我要实现的下一件事是删除当前卡住的背景。

我正在尝试通过grabCut()和GC_WITH_INIT_MASK来操纵grabcut.py以编程方式标记背景和前景，以获得更好的效果

import cv2 as cv
import numpy as np

img = cv.imread('./man.jpeg', cv.IMREAD_UNCHANGED)

gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)

profile = cv.CascadeClassifier('/usr/local/Cellar/opencv/4.1.0_2/share/opencv4/haarcascades/haarcascade_frontalface_default.xml')

faces = profile.detectMultiScale(gray, 1.3, 5)

print(len(faces))

if len(faces) > 1 or len(faces) == 0:
  print('more than one or zero found')
  exit()

for (x, y, w, h) in faces:
  y1 = int(y - h/2)
  y2 = int(y + h*2)
  x1 = int(x - w/2)
  x2 = int(x + w*2)
  //TODO: check if dimensions outside original img
  cropped = img[y1:y2, x1:x2]

//estimate marking sections which are BG and FG
cv.line(cropped, (int(cropped.shape[0]/6), 0), (int(cropped.shape[0]/6), int(cropped.shape[1]/3)), (0), 5)
cv.line(cropped, (int(cropped.shape[0]), 0), (int(cropped.shape[0] - cropped.shape[0]/4), int(cropped.shape[1]/3)), (0), 5)
cv.line(cropped, (int(cropped.shape[0]/6), int(cropped.shape[1] - cropped.shape[1]/6)), (int(cropped.shape[0] - cropped.shape[0]/6*2), int(cropped.shape[1] - cropped.shape[1]/6)), (255), 5)

mask = np.zeros(cropped.shape[:2], np.uint8)

bgModel = np.zeros((1, 65), np.float64) 
fgModel = np.zeros((1, 65), np.float64)

//getting an error of incorrect indicies?
#mask[cropped == 0] = 0
#mask[cropped == 255] = 1

cv.grabCut(cropped, mask, None, bgModel, fgModel, 5, cv.GC_INIT_WITH_MASK)

//tried both...second from grabcut.py
#mask2 = np.where((mask==2) | (mask==0), 0, 1).astype('uint8')
mask2 = np.where((mask==1) + (mask==3), 255, 0).astype('uint8')

output = cv.bitwise_and(cropped, cropped, mask=mask2)

cv.imshow('img', output)
cv.waitKey(0)
cv.destroyAllWindows()

现在我正在 error: (-215:Assertion failed) !bgdSamples.empty() && !fgdSamples.empty() in function 'initGMMs'

这基本上是许多不同教程，试验和错误的集合。我将其作为学习经验，因此任何指导将不胜感激。预先感谢！

Answer 1

您可以尝试

gray = cv2.cvtColor(cropped, cv2.COLOR_BGR2GRAY)
mask[gray == 0] = 0
mask[gray == 255] = 1

我认为也许在cv.GC_INIT_WITH_MASK模式下面具不能为空

OpenCV可以检测到人脸，裁剪轮廓，以编程方式标记并使用grabCut删除背景

1 个答案: