Character Dataset (DSchar10k)

The dataset Dissected Scene Characters 10k (DSChars10k ) contains 10,256 dissected character images. All images are alphabets with 52 classes, i.e., 26 classes for uppercase (A-Z) and another 26 classes for lowercase (a-z) characters. All character images are camera captured and Roman script. The dimension of all dissected character images is unequal. This dataset contains only 52 classes(alphabets only). All individual classes have an uneven number of samples.