TextZoom is a super-resolution dataset consisting of pairs of low-resolution (LR) and high-resolution (HR) images. These images are taken in the wild by cameras with different focal lengths. Each image pair is annotated with a case-sensitive string (including punctuation), bounding box type, and original focal length.

The dataset is divided into three subsets according to difficulty: difficult, medium, and easy. LR images are more challenging than synthetic LR images (BICUBIC) and can be used to study the field of low-resolution text recognition.