The Letter dataset, also known as the Letter Recognition Data Set, is a handwritten letter dataset. By randomly deforming pixel images of 26 uppercase letters in 20 commercial fonts, 20,000 unique character images are generated. The task of this dataset is mainly to recognize the 26 uppercase English letters from these images.