def write_to_file(self, f): """Write a BatchFile to a file object. f -- the file to write to (needs to support write) """ write_record(f, RecordTypes.BESTANDSVOORLOOP, **self.recordargs) for i, batch in enumerate(self.batches): batch.write_to_file(f, i+1) # i+1 -- people start counting at 1 write_record(f, RecordTypes.BESTANDSSLUIT)
def vis_one(vis_graph, filepath, samples, predictions, sv): movefile(filepath) record.write_record(None, reader) with sv.managed_session('', start_standard_services=False) as sess: sv.saver.restore(sess, tf.train.latest_checkpoint(A_checkpoint_dir)) a = sv.start_queue_runners(sess) save_dir = os.path.join(work_dir, _SEMANTIC_PREDICTION_SAVE_FOLDER) my_process_batch(sess, samples[common.ORIGINAL_IMAGE], predictions, samples[common.IMAGE_NAME], samples[common.HEIGHT], samples[common.WIDTH], save_dir)
def serve(filepath): global reader_graph movefile(filepath) # tf.reset_default_graph() get_reader() with reader_graph.as_default(): record.write_record(None, get_reader()) # input('record finish') vis_graph = tf.Graph() vis_graph.as_default() # with vis_graph.as_default(): # ------------------------------------------- samples, predictions = do_prepare() tf.train.get_or_create_global_step() saver = tf.train.Saver(slim.get_variables_to_restore()) # supervisor = tf.train.Supervisor(graph=vis_graph, supervisor = tf.train.Supervisor(init_op=tf.global_variables_initializer(), summary_op=None, summary_writer=None, global_step=None, saver=saver) with supervisor.managed_session(A_master, start_standard_services=False) as sess: supervisor.start_queue_runners(sess) my_checkpoint = tf.train.latest_checkpoint(A_checkpoint_dir) supervisor.saver.restore(sess, my_checkpoint) print(my_checkpoint) do_process_batch(sess, samples, predictions) skel_extract.extract() skel_extract.load() # graph.finalize() vis_graph.finalize()
def write_to_file(self, f, index): """Write a Batch to a file object. f -- the file to write to (needs to support write) index -- index number of this batch """ recordargs = self.recordargs recordargs.update({ 'batchvolgnummer': index, 'totaalbedrag': 0, 'totaalrekeningen': 0, 'aantalposten': 0, }) write_record(f, RecordTypes.BATCHVOORLOOP, **recordargs) try: self.description.write_to_file(f) except AttributeError: pass write_record(f, RecordTypes.OPDRACHTGEVER, **recordargs) # TODO: Loop over transactions write_record(f, RecordTypes.BATCHSLUIT, **recordargs)
def process_one(filepath): global reader_graph global vis_graph movefile(filepath) with reader_graph.as_default(): record.write_record(None, get_reader()) # with get_prepare_graph().as_default(): # samples, predictions = do_prepare() # # with vis_graph.as_default(): # tf.train.get_or_create_global_step() # sv = get_supervisor() # with sv.managed_session(A_master, start_standard_services=False) as sess: # do_process_batch(sess, samples, predictions) # skel_extract.extract() # skel_extract.load() with vis_graph.as_default(): with get_session().as_default(): samples, predictions = do_prepare() tf.train.get_or_create_global_step() do_process_batch(get_session(), samples, predictions)
def process_one(filepath): movefile(filepath) record.write_record(None, reader) # with get_prepare_graph().as_default(): # samples, predictions = do_prepare() # # with vis_graph.as_default(): # tf.train.get_or_create_global_step() # sv = get_supervisor() # with sv.managed_session(A_master, start_standard_services=False) as sess: # do_process_batch(sess, samples, predictions) # skel_extract.extract() # skel_extract.load() vis_graph = tf.Graph() with vis_graph.as_default(): dataset = segmentation_dataset.get_dataset(A_dataset, A_vis_split, dataset_dir=A_dataset_dir) samples = input_generator.get(dataset, A_vis_crop_size, A_vis_batch_size, min_resize_value=A_min_resize_value, max_resize_value=A_max_resize_value, resize_factor=A_resize_factor, dataset_split=A_vis_split, is_training=False, model_variant=A_model_variant) model_options = mycommon.ModelOptions( outputs_to_num_classes={common.OUTPUT_TYPE: dataset.num_classes}, crop_size=A_vis_crop_size, atrous_rates=A_atrous_rates, output_stride=A_output_stride) print(samples[common.IMAGE]) predictions = model.predict_labels(samples[common.IMAGE], model_options=model_options, image_pyramid=A_image_pyramid) predictions = predictions[common.OUTPUT_TYPE] tf.train.get_or_create_global_step() vis_session = tf.Session(graph=vis_graph) saver = tf.train.Saver(slim.get_variables_to_restore()) sv = tf.train.Supervisor(graph=vis_graph, logdir=A_vis_logdir, init_op=tf.global_variables_initializer(), summary_op=None, summary_writer=None, global_step=None, saver=saver) with sv.managed_session('', start_standard_services=False) as sess: sv.start_queue_runners(sess) sv.saver.restore(sess, tf.train.latest_checkpoint(A_checkpoint_dir)) #samples, predictions = do_prepare() #tf.train.get_or_create_global_step() #do_process_batch(get_session(), samples, predictions) save_dir = os.path.join(work_dir, _SEMANTIC_PREDICTION_SAVE_FOLDER) my_process_batch(sess, samples[common.ORIGINAL_IMAGE], predictions, samples[common.IMAGE_NAME], samples[common.HEIGHT], samples[common.WIDTH], save_dir)