ICDAR 2013 - Gender Identification Competition Dataset (GenderIdentifify2013)
Gender identification using all documents
In this task, one can include all the 4 documents of the writer for training purposes, this is the only task considered in the competition.
Researchers can also consider the following sub-tasks:
- Gender identification using Arabic handwriting documents only.
- Gender identification using English handwritings only.
- Text-dependent gender identification.
- Text-independent gender identification.
The evaluation metric used in the competition is the LogLoss metric. This evaluation requires the methods to output a probability value that each writer is male.
Other evaluation metrics can also be used including the Area Under the ROC Curve (AUC) as well as the recognition accuracy.
No comments on this dataset yet.