augment_data.py

import numpy as np
import os

# CONFIG
DATA_PATH = 'Alphabet_Data'
actions = np.array(['cool', 'who', 'computer']) # Add words you want to augment here!
# Or use os.listdir(DATA_PATH) to do ALL words

def add_noise(data, noise_factor=0.02):
    noise = np.random.normal(0, data.std() * noise_factor, data.shape)
    return data + noise

def scale_data(data, scale_factor=1.1):
    return data * scale_factor

for action in actions:
    action_path = os.path.join(DATA_PATH, action)
    if not os.path.exists(action_path): continue
    
    files = os.listdir(action_path)
    print(f"Augmenting {action}: Found {len(files)} sequences.")
    
    current_count = len(files)
    
    for i, file_name in enumerate(files):
        if not file_name.endswith('.npy'): continue
        
        # Load original
        data = np.load(os.path.join(action_path, file_name))
        
        # 1. Create "Jitter" version
        noisy_data = add_noise(data)
        np.save(os.path.join(action_path, str(current_count + i)), noisy_data)
        
        # 2. Create "Scaled" version
        scaled_data = scale_data(data)
        np.save(os.path.join(action_path, str(current_count + len(files) + i)), scaled_data)

    print(f" -> Done. New total: {len(os.listdir(action_path))} sequences.")