The
performance evaluation will
be based on a well established technique for writer identification.
More specifically, we plan to calculate the number of documents that
were assigned to the correct writer in terms of TOP-N choice. The
participant will be asked to provide a console application that will
produce the distance between two handwritten document images. For every
document in the benchmarking dataset, the distance among all documents
in this dataset will be calculated using the tool. A ranked list will
be produced and the evaluation will record the accuracy of the
participant's methodology in terms of TOP-1, TOP-10 and TOP-N choice (N
is the number of handwritten document images created by the same
writer).
