forked from ntu-rris/google-mediapipe
-
Notifications
You must be signed in to change notification settings - Fork 0
/
04_hand_rom.py
83 lines (67 loc) · 2.78 KB
/
04_hand_rom.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
###############################################################################
### Simple demo on measuring hand rom
### Input : Live video of hand
### Output: 2D display of hand keypoint
### with hand rom classification and corresponding joint angle
### Usage : python 04_hand_rom.py -m train (to log data)
### : python 04_hand_rom.py -m eval (to perform hand rom recognition)
###############################################################################
import cv2
import argparse
from utils_display import DisplayHand
from utils_mediapipe import MediaPipeHand
from utils_joint_angle import HandRomRecognition
parser = argparse.ArgumentParser()
parser.add_argument('-m', '--mode', default='eval', help='train / eval')
args = parser.parse_args()
mode = args.mode
# Load mediapipe hand class
pipe = MediaPipeHand(static_image_mode=False, max_num_hands=1)
# Load display class
disp = DisplayHand(max_num_hands=1)
# Start video capture
cap = cv2.VideoCapture(0) # By default webcam is index 0
# Load hand rom recognition class
gest = HandRomRecognition(mode)
counter = 0
while cap.isOpened():
ret, img = cap.read()
if not ret:
break
# Flip image for 3rd person view
img = cv2.flip(img, 1)
# To improve performance, optionally mark image as not writeable to pass by reference
img.flags.writeable = False
# Feedforward to extract keypoint
param = pipe.forward(img)
if (param[0]['class'] is not None) and (mode=='eval'):
param[0]['gesture'] = gest.eval(param[0]['angle'])
img.flags.writeable = True
# Display keypoint
cv2.imshow('img 2D', disp.draw2d(img.copy(), param))
key = cv2.waitKey(1)
if key==27:
break
if key==32 and (param[0]['class'] is not None) and (mode=='train'):
# Press spacebar to log training data
# Note: Need to manually change class label
# 'Finger MCP Flexion' :0,
# 'Finger PIP DIP Flexion':1,
# 'Thumb MCP Flexion' :2,
# 'Thumb IP Flexion' :3,
# 'Thumb Radial Abduction':4,
# 'Thumb Palmar Abduction':5, # From this class onwards hard to differentiate
# 'Thumb Opposition' :6,
# 'Forearm Pronation' :7, # Not done yet
# 'Forearm Supination' :8, # Not done yet
# 'Wrist Flexion' :9, # Not done yet
# 'Wrist Extension' :10,# Not done yet
# 'Wrist Radial Deviation':11,# Not done yet
# 'Wrist Ulnar Deviation' :12,# Not done yet
gest.train(param[0]['angle'], gest.gesture['Finger MCP Flexion'])
print('Saved', counter) # Log around 10 for each class
counter += 1
if key==32 and (param[0]['class'] is not None) and (mode=='eval'):
cv2.waitKey(0) # Pause display until user press any key
pipe.pipe.close()
cap.release()