Merge pull request #6 from kookmin-sw/feature/skin_detect

Feature/skin detect

.gitignore

@@ -0,0 +1 @@
+.DS_store

Skin_detect/skin_detect.py

@@ -0,0 +1,125 @@
+import dlib
+import cv2
+import numpy as np
+
+predictor_file = './shape_predictor_68_face_landmarks.dat' 
+image_file = './test/kimmingyu.jpeg' 
+
+ALL = list(range(0, 68))
+
+img = cv2.imread(image_file)
+
+# 얼굴의 landmark 뽑아내는 함수
+def landmark_extract(img, predictor_file):    
+    detector = dlib.get_frontal_face_detector()
+    predictor = dlib.shape_predictor(predictor_file)
+
+    gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    rects = detector(gray_img, 1)
+
+    for (i, rect) in enumerate(rects):
+        points = np.matrix([[p.x, p.y] for p in predictor(gray_img, rect).parts()])
+
+        left, right, top, bottom = rect.left(), rect.right(), rect.top(), rect.bottom()
+
+    return points, left, right, top, bottom
+
+# pos에 landmark 위치 정보 담겨있음
+RIGHT_EYE = list(range(36, 42))
+LEFT_EYE = list(range(42, 48))
+MOUTH = list(range(48, 68))
+NOSE = list(range(27, 36))
+EYEBROWS = list(range(17, 27))
+JAWLINE = list(range(1, 17))
+ALL = list(range(0, 68))
+EYES = list(range(36, 48))
+
+# 두 점 거리 찾기
+def find_distance(left, right):
+    a = np.abs(left[0] - right[0])
+    b = np.abs(left[1] - right[1])
+    return int(np.sqrt(np.power(a, 2) + np.power(b, 2)))
+
+# 두 점 기울기
+def find_slope(left, right):
+    x_increment = right[0] - left[0]
+    y_increment = right[1] - left[1]
+    return x_increment / y_increment
+
+# 두 점의 중심
+def find_center(left, right):
+    return (int((left[0] + right[0]) / 2), int((left[1] + right[1]) / 2)) 
+
+# 눈 부분 추출하기
+def eye_ousting(pos, img):
+    left_eye_pos = [(x[0], x[1]) for x in pos[LEFT_EYE].tolist()]
+    right_eye_pos = [(x[0], x[1]) for x in pos[RIGHT_EYE].tolist()]
+
+    # le_width = find_distance(left_eye_pos[0], left_eye_pos[3])
+    # ri_width = find_distance(right_eye_pos[0], right_eye_pos[3])
+
+    # le_height = find_distance(left_eye_pos[1], left_eye_pos[-2])
+    # ri_height = find_distance(right_eye_pos[1], right_eye_pos[-2])
+
+    # le_center = find_center(left_eye_pos[0], left_eye_pos[3])
+    # ri_center = find_center(right_eye_pos[0], right_eye_pos[3])
+
+    # left_slope = find_slope(left_eye_pos[0], left_eye_pos[3])
+    # right_slope = find_slope(right_eye_pos[0], right_eye_pos[3])
+
+    # left_axes_len = (le_width, le_height)
+    # right_axes_len = (ri_width, ri_height)
+
+    # mask = np.zeros_like(img[:, :])
+    # cv2.ellipse(img=mask, center=le_center, axes=left_axes_len, angle=left_slope, startAngle=0, endAngle=360, color=black, thickness=-1)
+    # cv2.ellipse(img=mask, center=ri_center, axes=right_axes_len, angle=right_slope, startAngle=0, endAngle=360, color=black, thickness=-1)
+
+    left_pts = np.array(left_eye_pos, np.int32)
+    left_pts = left_pts.reshape((-1, 1, 2))
+
+    right_pts = np.array(right_eye_pos, np.int32)
+    right_pts = right_pts.reshape((-1, 1, 2))
+
+    res = img.copy()
+
+    cv2.fillPoly(res, [left_pts], color=(0,0,0))
+    cv2.fillPoly(res, [right_pts], color=(0,0,0))
+
+    # arr = np.array([0, 0, 0], dtype=np.uint8)
+    # mask = cv2.inRange(mask, arr, arr)
+    # res = cv2.bitwise_and(img, img, mask=mask)
+
+    return res
+
+
+pos, left, right, top, bottom = landmark_extract(img, predictor_file)
+
+eye_ousting_img = eye_ousting(pos, img)
+
+face_original_img = img[top : bottom, left:right]
+face_img = eye_ousting_img[top : bottom, left:right]
+
+
+# https://www.researchgate.net/publication/262371199_Explicit_image_detection_using_YCbCr_space_color_model_as_skin_detection
+# lower
+lower = np.array([0, 133, 77], dtype=np.uint8)
+
+# upper
+upper = np.array([255, 173, 127], dtype=np.uint8)
+
+
+face_image_ycbcr = cv2.cvtColor(face_img, cv2.COLOR_BGR2YCrCb)
+face_skin_mask = cv2.inRange(face_image_ycbcr, lower, upper)
+
+# cv2.imshow("test", face_skin_mask)
+
+
+face_skin_img = cv2.bitwise_and(face_img, face_img, mask=face_skin_mask)
+
+image_horizontal = np.hstack((face_original_img, face_skin_img))
+
+cv2.imshow("Face Landmark", image_horizontal)
+key = cv2.waitKey(0)
+
+if key == ord('q'):
+    cv2.destroyAllWindows()

Skin_detect/skin_detect_v3.py

@@ -0,0 +1,32 @@
+import numpy as np
+import cv2
+
+img_path = "./test/kimmingyu.jpeg"
+
+img = cv2.imread(img_path)
+
+# convert BGR to hsv
+img_hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
+
+mask_hsv = cv2.inRange(img_hsv, (0, 15, 0), (17, 170, 255))
+mask_hsv = cv2.morphologyEx(mask_hsv, cv2.MORPH_OPEN, np.ones((3, 3), np.uint8))
+
+img_YCrCb = cv2.cvtColor(img, cv2.COLOR_BGR2YCrCb)
+
+mask_YCrCb = cv2.inRange(img_YCrCb, (0, 133, 77), (255, 173, 127))
+mask_YCrCb = cv2.morphologyEx(mask_YCrCb, cv2.MORPH_OPEN, np.ones((3, 3), np.uint8))
+
+mask = cv2.bitwise_and(mask_YCrCb, mask_hsv)
+mask = cv2.medianBlur(mask, 3)
+mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, np.ones((4, 4), np.uint8))
+
+res = cv2.bitwise_and(img, img, mask=mask)
+
+res_and_origin = np.hstack((res, img))
+
+cv2.imshow("res_and_origin", res_and_origin)
+
+key = cv2.waitKey(0)
+
+if key == ord('q'):
+    cv2.destroyAllWindows()

tempCodeRunnerFile.py

@@ -0,0 +1 @@
+