Back-Hand-Pose/data_convert.py at master · erwinwu211/Back-Hand-Pose · GitHub

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
import glob
import cv2
import numpy as np
import os
import math
import vg

data_path = 'data/BHP_Raw/TapData/P1'
window_size = 5
validation_split = 0.8

def preprocess_angle(y):

    temp = y[:,4,:]
    temp = temp.reshape(-1,25,3)
    #temp = temp[:,5:,:]
    y=[]
    for f in temp:
        #thumb
        thumb_v = f[4] - f[0]
        thumb_v = thumb_v / np.linalg.norm(thumb_v)
        f = f[5:,:]
        for i in range(4): #index~pinky finger
            for j in range(1,4):
                v1 = f[5*i+j] - f[5*i+j-1] #get previous bone vector
                v2 = f[5*i+j+1] - f[5*i+j] #get current bone vector
                if j == 1:  #MCP
                    #angle = vg.signed_angle(v1, v2, look=np.array([-1,0,0]))
                    angle = vg.signed_angle(v1, v2, look=np.array([0,-1,0]))
                    y.append(angle)
                angle = vg.signed_angle(v1, v2, look=np.array([-1,0,0]))
                y.append(angle)
        for i in thumb_v:
            y.append(i)
    y = np.array(y).reshape(-1,19)
    return y

def main():

    label_list = sorted(glob.glob(data_path+"/3d/*.npy"))

    for ind, l in enumerate(label_list):
        labels = np.load(l)[-300:]
        _, f = os.path.split(l)
        sess_name = f[:-8]
        img_dir = data_path + '/frames/' + sess_name
        img_list = sorted(glob.glob(img_dir+'/*.png'))
        imgs = []
        for i in img_list:
            frame = cv2.imread(i,cv2.IMREAD_UNCHANGED)
            frame = cv2.resize(frame,(224,224))
            imgs.append(frame)
        imgs = np.array(imgs)
        imgs = imgs.reshape(-1,window_size,  imgs.shape[1],imgs.shape[2],imgs.shape[3])
        labels = labels.reshape(-1,window_size,labels.shape[1])
        if ind == 0:
            x = imgs
            y = labels
        else:
            x = np.concatenate((x,imgs))
            y = np.concatenate((y,labels))
        print(ind,x.shape,y.shape)

    final_y = preprocess_angle(y)
    train_x = x[:int(len(x)*validation_split)]/255.0
    train_y = final_y[:int(len(final_y)*validation_split)]/90.0
    test_x = x[int(len(x)*validation_split):]/255.0
    test_y = final_y[int(len(final_y)*validation_split):]/90.0

    print(train_x.shape,train_y.shape)
    print(test_x.shape,test_y.shape)

    cond_name = data_path.split('/')[-2]
    usr_name = data_path.split('/')[-1]

    np.savez('data/train/'+cond_name +'_'+usr_name+'_window5.npz',train_x,train_y)
    np.savez('data/test/'+cond_name +'_'+usr_name+'_window5.npz',test_x,test_y)

main()