Major reductions made to the memory required by process_batch_list() so it can now process paragraph length documents from the CSAFE Handwriting Database on machines with 8 GB of RAM.
process_batch_list() now skips to the next document if an error is returned while trying to process a document. A log file records the document name(s) and error message(s) of any problem documents. If the user reruns process_batch_list() they now have the option to either try a second time to process problem documents or skip them entirely.
New function get_clusters_batch() calculates cluster assignments of all files in a directory.
fit_model() allows the user to fit a statistical model to known handwriting samples collected from a closed-set of persons.
analyze_questioned_documents() in conjunction with a model created by fit_model() allows a user to calculate the posterior probability that each known writer in the model is the writer of the questioned document(s). analyze_questioned_documents() only works when the questioned document(s) had to have been written by one of the model writers. This function must NOT be used if someone other than one of the model writers could written the questioned document(s) as it could yield misleading results.
processDocument() is a new wrapper function that runs readPNGBinary(), thinImage(), and processHandwriting() so the user doesn’t need to run these functions individually.
plotImage(), plotImageThinned(), and plotNodes now only need one input, a document processed with processDocument()
processHandwriting() no longer crashes when the input writing is a single letter or word
read_and_process() superseded in favor of processDocument()
extractGraphs() superseded in favor of process_batch_dir()