forked from handopen/hand-keypoint-detect
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdetect_on_camera.py
218 lines (201 loc) · 8.41 KB
/
detect_on_camera.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
import platform
sysstr = platform.system()
if (sysstr == "Windows"):
print("Call Windows tasks")
elif (sysstr == "Linux"): # 树莓派也是Linux
print("Call Linux tasks")
else:
print("other System tasks")
import copy
import time
import numpy as np
import cv2
from collections import deque
from hand_tracker import HandTracker
from gasture_utils.determine_gasture import create_known_finger_poses, determine_position, get_position_name_with_pose_id
from gasture_utils.FingerPoseEstimate import FingerPoseEstimate
"""
mediapipe 模型 handdetect模型
"""
palm_model_path = "./models/palm_detection_without_custom_op.tflite"
landmark_model_path = "./models/mediapipe_hand_landmark.tflite"
anchors_path = "./data/anchors.csv"
MIN_CONFIDENCE = 0.10
# cap = cv2.VideoCapture(r'C:\PythonProject\jing_vision\detection\keras_tf\tflite\posenet_mb2_ssd\dance.flv')
# cap = cv2.VideoCapture(r'hand.flv') # 使用本地视频
cap = cv2.VideoCapture(0) # 调用webcamera
cap.set(cv2.CAP_PROP_FPS, 30)
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 480)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)
detector = HandTracker(palm_model_path, landmark_model_path, anchors_path,
box_shift=0.2, box_enlarge=1.3)
Q = deque(maxlen=5)
gasture_define = { 'Simple Thumbs Up':0,
'Thumbs Up Right':1,
'I love you':2,
'Victory':3 ,
'Pointing Up':4,
'Okay':5 ,
'Spock':6,
'One':7,
'Three':8
}
reverse_gasture_define = {v:k for k,v in gasture_define.items()}
known_finger_poses = create_known_finger_poses()
# 8 12 16 20
# | | | |
# 7 11 15 19
# 4 | | | |
# | 6 10 14 18
# 3 | | | |
# | 5---9---13--17
# 2 \ /
# \ \ /
# 1 \ /
# \ \ /
# ------0-
# Anatomy guide
# http://blog.handcare.org/blog/2017/10/26/anatomy-101-finger-joints/
HAND_POINTS = [
"BASE",
"T_STT", "T_BCMC", "T_MCP", "T_IP", # Thumb
"I_CMC", "I_MCP", "I_PIP", "I_DIP", # Index
"M_CMC", "M_MCP", "M_PIP", "M_DIP", # Middle
"R_CMC", "R_MCP", "R_PIP", "R_DIP", # Ring
"P_CMC", "P_MCP", "P_PIP", "P_DIP", # Pinky
]
# 关节连接
limbs = [[0, 1],
[1, 2],
[2, 3],
[3, 4],
[0, 5],
[5, 6],
[6, 7],
[7, 8],
[0, 9],
[9, 10],
[10, 11],
[11, 12],
[0, 13],
[13, 14],
[14, 15],
[15, 16],
[0, 17],
[17, 18],
[18, 19],
[19, 20]
]
fps_list =[0 for i in range(10)]
joint_color_code = [[139, 53, 255],
[0, 56, 255],
[43, 140, 237],
[37, 168, 36],
[147, 147, 0],
[70, 17, 145]]
frame_count = 0
return_track={}
while True:
ret, color_image=cap.read()
show_image=copy.deepcopy(color_image)
color_image=cv2.cvtColor(color_image, cv2.COLOR_BGR2RGB) # 将BGR转为RGB
# print(ret, color_image.shape)
if not ret:
print('video over')
break
start_time=time.time()
# print(color_image.shape)
# 判断检测间隔
if frame_count%7==0:
kp, box,return_track=detector(color_image,None,True)
elif return_track is None:
pass
else:
kp, box, _ = detector(color_image,return_track,False)
if kp is not None and box is not None:
# print('kp is ',kp)
pts=np.array(box, np.int32)
kp=np.array(kp, np.int32)
# 计算手势
fingerPoseEstimate = FingerPoseEstimate(kp)
fingerPoseEstimate.calculate_positions_of_fingers(print_finger_info=False)
obtained_positions = determine_position(fingerPoseEstimate.finger_curled,
fingerPoseEstimate.finger_position, known_finger_poses,
0.45 * 10)
# print(obtained_positions)
# 根据字典的值进行排序
gasture_pre = sorted(obtained_positions.items(), key=lambda item: item[1], reverse=True)
# 仅绘制最高概率与绘制所有可能概率
if len(gasture_pre)>0:
Q.append(gasture_define[gasture_pre[0][0]]) # 通过队列中最多的元素输出为类别
counts = np.bincount(Q)
# 返回众数
max_index = np.argmax(counts)
print(reverse_gasture_define[max_index])
cv2.putText(show_image, 'rank 1 pre %s probably %f'%(reverse_gasture_define[max_index],gasture_pre[0][1]), (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 155), 1, cv2.LINE_AA)
# cv2.putText(show_image, 'rank 1 pre %s probably %f'%(gasture_pre[0][0],gasture_pre[0][1]), (30, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 155), 1, cv2.LINE_AA)
# for i,pre in enumerate(gasture_pre):
# cv2.putText(show_image, 'rank %d pre：%s probably%f'%(i+1,pre[0],pre[1]), (20, i*20+20), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 155), 1, cv2.LINE_AA)
for i in kp:
cv2.circle(show_image, center=(i[0], i[1]), radius=3, color=(0, 255, 0), thickness=-1)
# cv2.rectangle(show_image, (int(box[0,0]), int(box[0,1])), (int(box[1,0]), int(box[1,1])), (0, 255,255), 2)
# 绘制关键点链接线
cv2.line(show_image, (kp[0][0], kp[0][1]),
(kp[1][0], kp[1][1]), (255, 0, 0), 2)
cv2.line(show_image, (kp[1][0], kp[1][1]),
(kp[2][0], kp[2][1]), (255, 0, 0), 2)
cv2.line(show_image, (kp[2][0], kp[2][1]),
(kp[3][0], kp[3][1]), (255, 0, 0), 2)
cv2.line(show_image, (kp[3][0], kp[3][1]),
(kp[4][0], kp[4][1]), (255, 0, 0), 2)
cv2.line(show_image, (kp[0][0], kp[0][1]),
(kp[5][0], kp[5][1]), (255, 255, 0), 2)
cv2.line(show_image, (kp[5][0], kp[5][1]),
(kp[6][0], kp[6][1]), (255, 255, 0), 2)
cv2.line(show_image, (kp[6][0], kp[6][1]),
(kp[7][0], kp[7][1]), (255, 255, 0), 2)
cv2.line(show_image, (kp[7][0], kp[7][1]),
(kp[8][0], kp[8][1]), (255, 255, 0), 2)
cv2.line(show_image, (kp[5][0], kp[5][1]),
(kp[9][0], kp[9][1]), (0, 0, 255), 2)
cv2.line(show_image, (kp[9][0], kp[9][1]),
(kp[10][0], kp[10][1]), (0, 0, 255), 2)
cv2.line(show_image, (kp[10][0], kp[10][1]),
(kp[11][0], kp[11][1]), (0, 0, 255), 2)
cv2.line(show_image, (kp[11][0], kp[11][1]),
(kp[12][0], kp[12][1]), (0, 0, 255), 2)
cv2.line(show_image, (kp[9][0], kp[9][1]),
(kp[13][0], kp[13][1]), (255, 0, 255), 2)
cv2.line(show_image, (kp[13][0], kp[13][1]),
(kp[14][0], kp[14][1]), (255, 0, 255), 2)
cv2.line(show_image, (kp[14][0], kp[14][1]),
(kp[15][0], kp[15][1]), (255, 0, 255), 2)
cv2.line(show_image, (kp[15][0], kp[15][1]),
(kp[16][0], kp[16][1]), (255, 0, 255), 2)
cv2.line(show_image, (kp[0][0], kp[0][1]),
(kp[17][0], kp[17][1]), (0, 255, 0), 2)
cv2.line(show_image, (kp[13][0], kp[13][1]),
(kp[17][0], kp[17][1]), (0, 255, 0), 2)
cv2.line(show_image, (kp[17][0], kp[17][1]),
(kp[18][0], kp[18][1]), (0, 255, 0), 2)
cv2.line(show_image, (kp[18][0], kp[18][1]),
(kp[19][0], kp[19][1]), (0, 255, 0), 2)
cv2.line(show_image, (kp[19][0], kp[19][1]),
(kp[20][0], kp[20][1]), (0, 255, 0), 2)
cv2.polylines(show_image, [pts], True, (0, 255, 255), 1) # 绘制多边形
# print(box)
fps = 1 / (time.time() - start_time)
print('FPS:', fps)
index = frame_count%10
fps_list[index]=fps
cv2.putText(show_image,
'FPS:%f' % ( np.mean(fps_list)), (30, 70),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 155), 1, cv2.LINE_AA)
if sysstr == "Windows":
cv2.namedWindow("USB Camera", cv2.WINDOW_AUTOSIZE)
cv2.imshow("USB Camera", show_image)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
frame_count +=1
cap.release()
cv2.destroyAllWindows()