AFHQ stands for Animal Faces-HQ, which is an animal face dataset. The dataset consists of 15,000 high-quality images with a resolution of 512×512 pixels, including three fields: cats, dogs, and wild animals, with 5,000 images in each field.

This dataset sets a more challenging image translation task due to the presence of images from multiple fields (three) and multiple breeds (each field contains greater than or equal to eight breeds). All images are aligned vertically and horizontally to keep the eye in the center. Low-quality images have been manually removed.