def read_flags(): flags.DEFINE_integer("num_proposals", 3, "How many recognation proposals are shown") flags.DEFINE_string( "deactivate_touchscreen", "", # e.g. maXTouch Digitizer "When given this touchscreen is deactivated (for name see xinput)") dafault_model_path = os.path.join(os.path.dirname(__file__), 'data', 'model', 'model.ckpt') flags.DEFINE_string("model_path", dafault_model_path, "path of the model") return flags.FLAGS
def read_flags(): flags.DEFINE_string( "show_record", "", "if a path is provided, it shows this tfrecord") flags.DEFINE_boolean( "load_cache", False, "load all strokes from the pkl file from previous run") flags.DEFINE_string( "path", "tmp", "folder to save the records") flags.DEFINE_integer( "max_files", 10**10, "the max number of files to read") flags.DEFINE_string("my_handwriting_train_path", "data/my_handwriting/train/", "path folder with own examples") flags.DEFINE_string("my_handwriting_test_path", "data/my_handwriting/test/", "path folder with own examples") return flags.FLAGS
def read_flags(): flags.DEFINE_string('restore_from', "", "optional ckpt-file to restore-from") flags.DEFINE_string('path', "vanilla", "Name of the network") flags.DEFINE_string('lstm_sizes', "[120, 120]", "List of LSTM-layer sizes (json)") flags.DEFINE_boolean( 'share_param_first_layer', True, "Whether the first LSTM-layer shares parameters" "in forward and backward layer") flags.DEFINE_integer('epoch_num', 1, "Number of epochs to train") flags.DEFINE_integer('steps_per_checkpoint', 100, "") #flags.DEFINE_integer( # 'num_steps', 10000, "Number of batches to run") flags.DEFINE_float('learning_rate', 0.003, "Learning rate for Optimizer") flags.DEFINE_boolean('fine_tuning', True, "reduce learning rate to 20% for last epoch") #flags.DEFINE_integer( # 'num_final_steps', 1000, # "Number of batches to run before learning_rate_fine is used") flags.DEFINE_string('data_path', "./records", "filepath to the tfrecords") flags.DEFINE_boolean('batch_size', 32, "Size of the training batch") return flags.FLAGS
You should have received a copy of the GNU General Public License along with Smart Manuscript. If not, see <http://www.gnu.org/licenses/>. """ from tensorflow.python.platform.app import flags import os import glob from .inkml import InkML, TraceGroup from .corpus import Corpus, Corpora, TranscriptedStrokes __author__ = "Daniel Vorberg" __copyright__ = "Copyright (c) 2017, Daniel Vorberg" __license__ = "GPL" flags.DEFINE_string("ibm_ub_path", "data/IBM_UB_1/query/", "path to IBM_UB_1 folder (unzipped)") class IBMub(InkML): TRANSCRIPTION = "Tg_Truth" def get_segments(self): for trace_group in self._root.childs_of_type(TraceGroup): yield TranscriptedStrokes( transcription=trace_group.annotation[self.TRANSCRIPTION], strokes=trace_group.ink()) def load(path, max_files=None): """ return corpora (corpus of all files) of the IBM UB database
return svg_path def _read_pdf(filename, is_handwritten=None): svg_filename = _pdf_to_svg_tmp(filename) return _read_svg(svg_filename, is_handwritten) def main(svg_filename): """ show sample handwritten notes """ strokes, page_size = load(svg_filename) ink = InkPage(strokes, page_size) ink.plot_pylab() plt.show() for line in ink.lines: line.plot_pylab() plt.show() if __name__ == "__main__": from tensorflow.python.platform.app import flags FLAGS = flags.FLAGS flags.DEFINE_string( "file", os.path.join(os.path.dirname(__file__), "data", "sample_text", "The_Zen_of_Python.pdf"), "file to show features (either PDF or SVG)") main(FLAGS.file)
""" from tensorflow.python.platform.app import flags import pylab as plt import os import glob from copy import deepcopy from .corpus import Corpus, Corpora, TranscriptedStrokes from .inkml import InkML, TraceView __author__ = "Daniel Vorberg" __copyright__ = "Copyright (c) 2017, Daniel Vorberg" __license__ = "GPL" flags.DEFINE_string("iam_on_do_path", "data/IAMonDo-db-1.0", "path to IAMonDo-db-1.0 folder (unzipped)") class IAMonDo(InkML): TEXTLINE = "Textline" WORD = "Word" TRANSCRIPTION = "transcription" TYPE = "type" CORRECTION = "Correction" @staticmethod def condition(element, type_): def includes_corrections(element): child_include_corrections = any( next(child.search(includes_corrections), None) is not None