Download and unzip https://github.com/myleott/mnist_png/blob/master/mnist_png.tar.gz
Make sure vanilla mnist data is at ./mnist_png/training/
Run python prepare_dataset.py ./mnist_png/training/ ./watermark_mnist_png_random/training
Watermarked data will be placed under ./watermark_mnist_png/training
Example output: https://drive.google.com/file/d/0B7dTwsaPqx5nZ2xLVEFLdXlvY2c/view?usp=sharing
Download and unzip https://drive.google.com/open?id=0B-ujUXlVFw1zajF0eThCYjhYc0k
Make sure vanilla cifar data is at ./cifar_png/training/
Run python prepare_dataset.py ./cifar_png/training/ ./cifar_mnist_png_random/training
Watermarked data will be placed under ./watermark_cifar_png/training