ability to split out train/test/valid folders on export

[garminTestPilot50]

For yolo training, it is super useful to be able to export out the train/test/valid folders and be able to control the mix (like 80% / 10% / 10%) for the images, and randomly assign the images & labels to those folders with proper pairing.

I ended up writing some python code to move the images & labels afterwards, but it would be great to be able to do that from the tool itself.

Here is the simple code if it helps:

import glob
import random
import os
import shutil

Use raw strings for Windows file paths

PATH = r'F:\my\folder\data\'
img_paths = glob.glob(PATH + r'images*.jpg')
txt_paths = glob.glob(PATH + r'labels*.txt')

Calculate number of files for training, validation

data_size = len(img_paths)
train_size = int(data_size * 0.8)
test_size = int(data_size * 0.1)
valid_size = data_size - train_size - test_size # Ensure all data is used

Shuffle two lists

img_txt = list(zip(img_paths, txt_paths))
random.seed(43)
random.shuffle(img_txt)
img_paths, txt_paths = zip(*img_txt)

Now split them

train_img_paths = img_paths[:train_size]
train_txt_paths = txt_paths[:train_size]

test_img_paths = img_paths[train_size:train_size + test_size]
test_txt_paths = txt_paths[train_size:train_size + test_size]

valid_img_paths = img_paths[train_size + test_size:]
valid_txt_paths = txt_paths[train_size + test_size:]

Function to create folders and move files

def create_and_move_files(folder_path, img_paths, txt_paths):
if not os.path.exists(folder_path):
os.makedirs(folder_path, exist_ok=True) # Use makedirs to create intermediate directories if needed
for img_path, txt_path in zip(img_paths, txt_paths):
try:
shutil.move(img_path, folder_path)
shutil.move(txt_path, folder_path)
except Exception as e:
print(f"An error occurred while moving files: {e}")

Move them to train, valid, test folders

create_and_move_files(PATH + r'train\', train_img_paths, train_txt_paths)
create_and_move_files(PATH + r'test\', test_img_paths, test_txt_paths)
create_and_move_files(PATH + r'valid\', valid_img_paths, valid_txt_paths)