ML-Motivators commited on
Commit
0ae2801
·
verified ·
1 Parent(s): 4dd837f

Delete utils_mask.py

Browse files
Files changed (1) hide show
  1. utils_mask.py +0 -167
utils_mask.py DELETED
@@ -1,167 +0,0 @@
1
- import numpy as np
2
- import cv2
3
- from PIL import Image, ImageDraw
4
-
5
- label_map = {
6
- "background": 0,
7
- "hat": 1,
8
- "hair": 2,
9
- "sunglasses": 3,
10
- "upper_clothes": 4,
11
- "skirt": 5,
12
- "pants": 6,
13
- "dress": 7,
14
- "belt": 8,
15
- "left_shoe": 9,
16
- "right_shoe": 10,
17
- "head": 11,
18
- "left_leg": 12,
19
- "right_leg": 13,
20
- "left_arm": 14,
21
- "right_arm": 15,
22
- "bag": 16,
23
- "scarf": 17,
24
- }
25
-
26
- def extend_arm_mask(wrist, elbow, scale):
27
- wrist = elbow + scale * (wrist - elbow)
28
- return wrist
29
-
30
- def hole_fill(img):
31
- img = np.pad(img[1:-1, 1:-1], pad_width = 1, mode = 'constant', constant_values=0)
32
- img_copy = img.copy()
33
- mask = np.zeros((img.shape[0] + 2, img.shape[1] + 2), dtype=np.uint8)
34
-
35
- cv2.floodFill(img, mask, (0, 0), 255)
36
- img_inverse = cv2.bitwise_not(img)
37
- dst = cv2.bitwise_or(img_copy, img_inverse)
38
- return dst
39
-
40
- def refine_mask(mask):
41
- contours, hierarchy = cv2.findContours(mask.astype(np.uint8),
42
- cv2.RETR_CCOMP, cv2.CHAIN_APPROX_TC89_L1)
43
- area = []
44
- for j in range(len(contours)):
45
- a_d = cv2.contourArea(contours[j], True)
46
- area.append(abs(a_d))
47
- refine_mask = np.zeros_like(mask).astype(np.uint8)
48
- if len(area) != 0:
49
- i = area.index(max(area))
50
- cv2.drawContours(refine_mask, contours, i, color=255, thickness=-1)
51
-
52
- return refine_mask
53
-
54
- def get_mask_location(model_type, category, model_parse: Image.Image, keypoint: dict, width=384,height=512):
55
- im_parse = model_parse.resize((width, height), Image.NEAREST)
56
- parse_array = np.array(im_parse)
57
-
58
- if model_type == 'hd':
59
- arm_width = 60
60
- elif model_type == 'dc':
61
- arm_width = 45
62
- else:
63
- raise ValueError("model_type must be \'hd\' or \'dc\'!")
64
-
65
- parse_head = (parse_array == 1).astype(np.float32) + \
66
- (parse_array == 3).astype(np.float32) + \
67
- (parse_array == 11).astype(np.float32)
68
-
69
- parser_mask_fixed = (parse_array == label_map["left_shoe"]).astype(np.float32) + \
70
- (parse_array == label_map["right_shoe"]).astype(np.float32) + \
71
- (parse_array == label_map["hat"]).astype(np.float32) + \
72
- (parse_array == label_map["sunglasses"]).astype(np.float32) + \
73
- (parse_array == label_map["bag"]).astype(np.float32)
74
-
75
- parser_mask_changeable = (parse_array == label_map["background"]).astype(np.float32)
76
-
77
- arms_left = (parse_array == 14).astype(np.float32)
78
- arms_right = (parse_array == 15).astype(np.float32)
79
-
80
- if category == 'dresses':
81
- parse_mask = (parse_array == 7).astype(np.float32) + \
82
- (parse_array == 4).astype(np.float32) + \
83
- (parse_array == 5).astype(np.float32) + \
84
- (parse_array == 6).astype(np.float32)
85
-
86
- parser_mask_changeable += np.logical_and(parse_array, np.logical_not(parser_mask_fixed))
87
-
88
- elif category == 'upper_body':
89
- parse_mask = (parse_array == 4).astype(np.float32) + (parse_array == 7).astype(np.float32)
90
- parser_mask_fixed_lower_cloth = (parse_array == label_map["skirt"]).astype(np.float32) + \
91
- (parse_array == label_map["pants"]).astype(np.float32)
92
- parser_mask_fixed += parser_mask_fixed_lower_cloth
93
- parser_mask_changeable += np.logical_and(parse_array, np.logical_not(parser_mask_fixed))
94
- elif category == 'lower_body':
95
- parse_mask = (parse_array == 6).astype(np.float32) + \
96
- (parse_array == 12).astype(np.float32) + \
97
- (parse_array == 13).astype(np.float32) + \
98
- (parse_array == 5).astype(np.float32)
99
- parser_mask_fixed += (parse_array == label_map["upper_clothes"]).astype(np.float32) + \
100
- (parse_array == 14).astype(np.float32) + \
101
- (parse_array == 15).astype(np.float32)
102
- parser_mask_changeable += np.logical_and(parse_array, np.logical_not(parser_mask_fixed))
103
- else:
104
- raise NotImplementedError
105
-
106
- # Load pose points
107
- pose_data = keypoint["pose_keypoints_2d"]
108
- pose_data = np.array(pose_data)
109
- pose_data = pose_data.reshape((-1, 2))
110
-
111
- im_arms_left = Image.new('L', (width, height))
112
- im_arms_right = Image.new('L', (width, height))
113
- arms_draw_left = ImageDraw.Draw(im_arms_left)
114
- arms_draw_right = ImageDraw.Draw(im_arms_right)
115
- if category == 'dresses' or category == 'upper_body':
116
- shoulder_right = np.multiply(tuple(pose_data[2][:2]), height / 512.0)
117
- shoulder_left = np.multiply(tuple(pose_data[5][:2]), height / 512.0)
118
- elbow_right = np.multiply(tuple(pose_data[3][:2]), height / 512.0)
119
- elbow_left = np.multiply(tuple(pose_data[6][:2]), height / 512.0)
120
- wrist_right = np.multiply(tuple(pose_data[4][:2]), height / 512.0)
121
- wrist_left = np.multiply(tuple(pose_data[7][:2]), height / 512.0)
122
- ARM_LINE_WIDTH = int(arm_width / 512 * height)
123
- size_left = [shoulder_left[0] - ARM_LINE_WIDTH // 2, shoulder_left[1] - ARM_LINE_WIDTH // 2, shoulder_left[0] + ARM_LINE_WIDTH // 2, shoulder_left[1] + ARM_LINE_WIDTH // 2]
124
- size_right = [shoulder_right[0] - ARM_LINE_WIDTH // 2, shoulder_right[1] - ARM_LINE_WIDTH // 2, shoulder_right[0] + ARM_LINE_WIDTH // 2,
125
- shoulder_right[1] + ARM_LINE_WIDTH // 2]
126
-
127
-
128
- if wrist_right[0] <= 1. and wrist_right[1] <= 1.:
129
- im_arms_right = arms_right
130
- else:
131
- wrist_right = extend_arm_mask(wrist_right, elbow_right, 1.2)
132
- arms_draw_right.line(np.concatenate((shoulder_right, elbow_right, wrist_right)).astype(np.uint16).tolist(), 'white', ARM_LINE_WIDTH, 'curve')
133
- arms_draw_right.arc(size_right, 0, 360, 'white', ARM_LINE_WIDTH // 2)
134
-
135
- if wrist_left[0] <= 1. and wrist_left[1] <= 1.:
136
- im_arms_left = arms_left
137
- else:
138
- wrist_left = extend_arm_mask(wrist_left, elbow_left, 1.2)
139
- arms_draw_left.line(np.concatenate((wrist_left, elbow_left, shoulder_left)).astype(np.uint16).tolist(), 'white', ARM_LINE_WIDTH, 'curve')
140
- arms_draw_left.arc(size_left, 0, 360, 'white', ARM_LINE_WIDTH // 2)
141
-
142
- hands_left = np.logical_and(np.logical_not(im_arms_left), arms_left)
143
- hands_right = np.logical_and(np.logical_not(im_arms_right), arms_right)
144
- parser_mask_fixed += hands_left + hands_right
145
-
146
- parser_mask_fixed = np.logical_or(parser_mask_fixed, parse_head)
147
- parse_mask = cv2.dilate(parse_mask, np.ones((5, 5), np.uint16), iterations=5)
148
- if category == 'dresses' or category == 'upper_body':
149
- neck_mask = (parse_array == 18).astype(np.float32)
150
- neck_mask = cv2.dilate(neck_mask, np.ones((5, 5), np.uint16), iterations=1)
151
- neck_mask = np.logical_and(neck_mask, np.logical_not(parse_head))
152
- parse_mask = np.logical_or(parse_mask, neck_mask)
153
- arm_mask = cv2.dilate(np.logical_or(im_arms_left, im_arms_right).astype('float32'), np.ones((5, 5), np.uint16), iterations=4)
154
- parse_mask += np.logical_or(parse_mask, arm_mask)
155
-
156
- parse_mask = np.logical_and(parser_mask_changeable, np.logical_not(parse_mask))
157
-
158
- parse_mask_total = np.logical_or(parse_mask, parser_mask_fixed)
159
- inpaint_mask = 1 - parse_mask_total
160
- img = np.where(inpaint_mask, 255, 0)
161
- dst = hole_fill(img.astype(np.uint8))
162
- dst = refine_mask(dst)
163
- inpaint_mask = dst / 255 * 1
164
- mask = Image.fromarray(inpaint_mask.astype(np.uint8) * 255)
165
- mask_gray = Image.fromarray(inpaint_mask.astype(np.uint8) * 127)
166
-
167
- return mask, mask_gray