FLK/prepare_data.py at main · PARCO-LAB/FLK · 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
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
import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler
import random

import warnings
warnings.filterwarnings("ignore")

import glob

path = 'data/aist/train/gt' # trained on ground truth


OP_H36M_NAMES = ['']*12

OP_H36M_NAMES[0]  = 'RHip'
OP_H36M_NAMES[1]  = 'RKnee'
OP_H36M_NAMES[2]  = 'RAnkle'
OP_H36M_NAMES[3]  = 'LHip'
OP_H36M_NAMES[4]  = 'LKnee'
OP_H36M_NAMES[5]  = 'LAnkle'
OP_H36M_NAMES[6]  = 'LShoulder'
OP_H36M_NAMES[7]  = 'LElbow'
OP_H36M_NAMES[8]  = 'LWrist'
OP_H36M_NAMES[9] = 'RShoulder'
OP_H36M_NAMES[10] = 'RElbow'
OP_H36M_NAMES[11] = 'RWrist'


def center_on_hip_first(df : pd.DataFrame, kp_center = 'Hip'):
    columns = df.columns
    df.reset_index(inplace = True)
    row = df.loc[0]

        #find MidHip
    RHip = [row['R' + kp_center +':X'],row['R' + kp_center +':Y'],row['R' + kp_center +':Z']]
    LHip = [row['L' + kp_center +':X'],row['L' + kp_center +':Y'],row['L' + kp_center +':Z']]
    MidHip = [(RHip[0] + LHip[0])/2, (RHip[1] + LHip[1])/2, (RHip[2] + LHip[2])/2]
    for i in range(len(df)):
        row = df.loc[i]
        for j in columns:
            if j.split(':')[-1] == 'X':
                row[j] = row[j] - MidHip[0]
            elif j.split(':')[-1] == 'Y':
                row[j] = row[j] - MidHip[1]
            elif j.split(':')[-1] == 'Z':
                row[j] = row[j] - MidHip[2]


        df.loc[i] = row

    return df

def load_from_csv(files : list, timesteps : int, scaler : StandardScaler = None, sample = None):

    final_X = []
    final_y = []
    for filename in files:
        print("processing",filename)
        df = pd.read_csv(filename)
        if 'time' in df:
            df = df.drop(columns = ['time'])
        X = []
        y = []
        columns_full = list(df.columns)
        to_drop = []
        for i in columns_full:
            if i.split(':')[0] not in OP_H36M_NAMES:
                to_drop.append(i)

        df = df.drop(columns = to_drop)
        if not scaler:
            for j in range(len(df) - timesteps):
                in_seq = df[j: j + timesteps].values.tolist()
                out = df.loc[j + timesteps].values.tolist()
                X.append(in_seq)
                y.append(out)
            X = np.array(X)
            y = np.array(y)
        else:
            df_values = df.values

            for j in range(len(df) - timesteps):
                sequenza = df[j: j + timesteps + 1]
                sequenza = center_on_hip_first(sequenza)
                sequenza = sequenza.drop(columns=['index'])
                s_mean = np.nanmean(sequenza)
                s_std = np.nanstd(sequenza)
                sequenza = (sequenza - s_mean)/s_std
                in_seq = sequenza[:-1].values.tolist()
                out = sequenza.values.tolist()[-1]
                X.append(in_seq)
                y.append(out)

            if len(X) > sample:
                subset_index = np.random.choice(list(range(0, len(X))),size = sample, replace = False)
                X = [X[i] for i in subset_index]
                y = [y[i] for i in subset_index]

            final_X += X
            print(np.array(final_X).shape)
            final_y += y

    subset_index = np.random.choice(list(range(0, len(final_X))),size = len(final_X), replace = False)
    final_X = [final_X[i] for i in subset_index]
    final_y = [final_y[i] for i in subset_index]
    X = np.array(final_X)
    y = np.array(final_y)


    split_index = int(len(X) * 0.8)  # 80% training, 20% validation
    X_train, X_val = X[:split_index], X[split_index:]
    y_train, y_val = y[:split_index], y[split_index:]

    if scaler:
        return X_train, X_val, y_train, y_val, scaler
    else:
        return X_train, X_val, y_train, y_val, None


scaler = StandardScaler()


files = glob.glob(path + '/*.csv')

#files = files[:74]
files_final=[]
for i in files:
    if 'S9' in i or 'S11' in i:
        pass
    else:
        files_final.append(i)

length = 64
_sample = 10
dataset = 'aist'
X_train, X_val, y_train, y_val, scaler = load_from_csv(files_final,length, scaler, sample = _sample)

np.save('dataset/training_FLK/' + dataset + '/X_train_' + str(length) + '_' + str(_sample)+  '.npy', X_train)
np.save('dataset/training_FLK/' + dataset + '/X_val_' + str(length) + '_' + str(_sample)+  '.npy', X_val)
np.save('dataset/training_FLK/' + dataset + '/y_train_' + str(length) + '_' + str(_sample)+  '.npy', y_train)
np.save('dataset/training_FLK/' + dataset + '/y_val_' + str(length) + '_' + str(_sample)+  '.npy', y_val)