preprocess_joints.py

import numpy as np
from hand_cropping import crop_hand_data
import os

image_data = []
label_data = []
for c in ['A','B','C','D','E']:
    for path in os.listdir(f'data/{c}'):
        if path == '.DS_Store':
            continue
        output = crop_hand_data(f'data/{c}/{path}')
        image_data.append(output)
        label_data.append([ord(path)-ord('a')]*output.shape[0])

image_data = np.concatenate(image_data)
label_data = np.concatenate(label_data)
indices = np.arange(len(label_data))
np.random.shuffle(indices)
image_data = image_data[indices]
label_data = label_data[indices]
np.save('data/n_joints',image_data)
np.save('data/n_labels',label_data)

# print(image_data)

#=======================================================================================================================================
#code from train_test_split.py

joints = np.load('data/n_joints.npy')
labels = np.load('data/n_labels.npy')

test_split = 0.2

train_len = int(len(joints)*(1-test_split))
train_joints = joints[:train_len]
test_joints = joints[train_len:]
train_labels = labels[:train_len]
test_labels = labels[train_len:]

np.save('data/n_joints_train',train_joints.astype(np.float32))
np.save('data/n_joints_test',test_joints.astype(np.float32))
np.save('data/n_labels_train',train_labels.astype(np.int64))
np.save('data/n_labels_test',test_labels.astype(np.int64))