This repository provides a demonstration of how to use our proposed ZS-N2N (Zero-Shot Noise2Noise) algorithm to denoise images without any training data or a noise model or level as input. The notebook is Pytorch based. The method is as simple as convolving the noisy image with two fixed kernels to create a pair of downsampled images. A simple 2 layer CNN is then trained with a consistency loss to map one downsampled image to the other. The notebook can easily be run on GPU or CPU. Choose the device to be 'cuda' if you want to run on GPU, otherwise choose 'cpu'. If you choose 'cuda' remember to change your colab runtime to GPU via 'Runtime' -> 'Change runtime type'.