AHD is a dataset of abnormal handwritten digits collected by disabled people. Additionally, a few digits were collected from people with normal physical conditions to cover these types of digits as well. The AHD consists of images from several handwritten sets, using different writing tools, including pencils, pens, markers, etc.
Dataset description and the English digits are available in the repo.
For more details, you can send an email to EshaghMoutabi95@gmail.com
In this dataset, we attempted to consider some specific characteristics of handwritten digits that were essential for solving problems because there was no suitable dataset. There are many types of abnormality, which creates various issues and reduces the accuracy of models. The dataset includes different abnormal English digits collected during a month and automatically pre-processed and augmented, along with some typical cases to increase the study's generality. We considered various factors in the dataset like:
The following are some samples and also a UMAP image of a scanned page.
Each rows belong to one person.
