PyTorch Image Quality (PIQ) is a collection of perceptual and full-reference metrics for assessing image quality, primarily used for evaluating image generation, restoration, and enhancement models. It implements common metrics like SSIM, PSNR, MS-SSIM, FID, and KID as PyTorch modules, many of which can be used as differentiable loss functions during training. The library aims to simplify model evaluation by providing a unified interface with validation, replacing boilerplate code for computing image quality metrics.