def create_train_dataset(args, dataset_args=None):
gt_extension = args.gt_extension if args.gt_extension is not None else DataSetType.gt_extension(args.dataset)
# Training dataset
print("Resolving input files")
input_image_files = sorted(glob_all(args.files))
if not args.text_files:
if gt_extension:
gt_txt_files = [split_all_ext(f)[0] + gt_extension for f in input_image_files]
else:
gt_txt_files = [None] * len(input_image_files)
else:
gt_txt_files = sorted(glob_all(args.text_files))
input_image_files, gt_txt_files = keep_files_with_same_file_name(input_image_files, gt_txt_files)
for img, gt in zip(input_image_files, gt_txt_files):
if split_all_ext(os.path.basename(img))[0] != split_all_ext(os.path.basename(gt))[0]:
raise Exception("Expected identical basenames of file: {} and {}".format(img, gt))
if len(set(gt_txt_files)) != len(gt_txt_files):
raise Exception("Some image are occurring more than once in the data set.")
dataset = create_dataset(
args.dataset,
DataSetMode.TRAIN,
images=input_image_files,
texts=gt_txt_files,
skip_invalid=not args.no_skip_invalid_gt,
args=dataset_args if dataset_args else {},
)
print("Found {} files in the dataset".format(len(dataset)))
return dataset
def create_test_dataset(
cfg: CfgNode,
dataset_args=None
) -> Union[List[Union[RawDataSet, FileDataSet, AbbyyDataSet, PageXMLDataset,
Hdf5DataSet, ExtendedPredictionDataSet,
GeneratedLineDataset]], None]:
if cfg.DATASET.VALID.TEXT_FILES:
assert len(cfg.DATASET.VALID.PATH) == len(cfg.DATASET.VALID.TEXT_FILES)
if cfg.DATASET.VALID.PATH:
validation_dataset_list = []
print("Resolving validation files")
for i, valid_path in enumerate(cfg.DATASET.VALID.PATH):
validation_image_files = glob_all(valid_path)
dataregistry.register(
i, os.path.basename(os.path.dirname(valid_path)),
len(validation_image_files))
if not cfg.DATASET.VALID.TEXT_FILES:
val_txt_files = [
split_all_ext(f)[0] + cfg.DATASET.VALID.GT_EXTENSION
for f in validation_image_files
]
else:
val_txt_files = sorted(
glob_all(cfg.DATASET.VALID.TEXT_FILES[i]))
validation_image_files, val_txt_files = keep_files_with_same_file_name(
validation_image_files, val_txt_files)
for img, gt in zip(validation_image_files, val_txt_files):
if split_all_ext(
os.path.basename(img))[0] != split_all_ext(
os.path.basename(gt))[0]:
raise Exception(
"Expected identical basenames of validation file: {} and {}"
.format(img, gt))
if len(set(val_txt_files)) != len(val_txt_files):
raise Exception(
"Some validation images are occurring more than once in the data set."
)
validation_dataset = create_dataset(
cfg.DATASET.VALID.TYPE,
DataSetMode.TRAIN,
images=validation_image_files,
texts=val_txt_files,
skip_invalid=not cfg.DATALOADER.NO_SKIP_INVALID_GT,
args=dataset_args,
)
print("Found {} files in the validation dataset".format(
len(validation_dataset)))
validation_dataset_list.append(validation_dataset)
else:
validation_dataset_list = None
return validation_dataset_list
def main():
parser = ArgumentParser()
parser.add_argument("--pred", nargs="+", required=True,
help="Extended prediction files (.json extension)")
args = parser.parse_args()
print("Resolving files")
pred_files = sorted(glob_all(args.pred))
data_set = create_dataset(
DataSetType.EXTENDED_PREDICTION,
DataSetMode.EVAL,
texts=pred_files,
)
print('Average confidence: {:.2%}'.format(np.mean([s['best_prediction'].avg_char_probability for s in data_set.samples()])))
def main():
parser = ArgumentParser()
parser.add_argument("--pred", nargs="+", required=True,
help="Extended prediction files (.json extension)")
args = parser.parse_args()
print("Resolving files")
pred_files = sorted(glob_all(args.pred))
data_set = create_dataset(
DataSetType.EXTENDED_PREDICTION,
DataSetMode.EVAL,
texts=pred_files,
)
data_set.load_samples(progress_bar=True)
print('Average confidence: {:.2%}'.format(np.mean([s['best_prediction'].avg_char_probability for s in data_set.samples()])))
def create_train_dataset(cfg: CfgNode, dataset_args=None):
gt_extension = cfg.DATASET.TRAIN.GT_EXTENSION if cfg.DATASET.TRAIN.GT_EXTENSION is not False else DataSetType.gt_extension(
cfg.DATASET.TRAIN.TYPE)
# Training dataset
print("Resolving input files")
input_image_files = sorted(glob_all(cfg.DATASET.TRAIN.PATH))
if not cfg.DATASET.TRAIN.TEXT_FILES:
if gt_extension:
gt_txt_files = [
split_all_ext(f)[0] + gt_extension for f in input_image_files
]
else:
gt_txt_files = [None] * len(input_image_files)
else:
gt_txt_files = sorted(glob_all(cfg.DATASET.TRAIN.TEXT_FILES))
input_image_files, gt_txt_files = keep_files_with_same_file_name(
input_image_files, gt_txt_files)
for img, gt in zip(input_image_files, gt_txt_files):
if split_all_ext(os.path.basename(img))[0] != split_all_ext(
os.path.basename(gt))[0]:
raise Exception(
"Expected identical basenames of file: {} and {}".format(
img, gt))
if len(set(gt_txt_files)) != len(gt_txt_files):
raise Exception(
"Some image are occurring more than once in the data set.")
dataset = create_dataset(
cfg.DATASET.TRAIN.TYPE,
DataSetMode.TRAIN,
images=input_image_files,
texts=gt_txt_files,
skip_invalid=not cfg.DATALOADER.NO_SKIP_INVALID_GT,
args=dataset_args if dataset_args else {},
)
print("Found {} files in the dataset".format(len(dataset)))
return dataset
def main():
parser = ArgumentParser()
parser.add_argument("--dataset",
type=DataSetType.from_string,
choices=list(DataSetType),
default=DataSetType.FILE)
parser.add_argument(
"--gt",
nargs="+",
required=True,
help="Ground truth files (.gt.txt extension). "
"Optionally, you can pass a single json file defining all parameters.")
parser.add_argument(
"--pred",
nargs="+",
default=None,
help=
"Prediction files if provided. Else files with .pred.txt are expected at the same "
"location as the gt.")
parser.add_argument("--pred_dataset",
type=DataSetType.from_string,
choices=list(DataSetType),
default=DataSetType.FILE)
parser.add_argument("--pred_ext",
type=str,
default=".pred.txt",
help="Extension of the predicted text files")
parser.add_argument(
"--n_confusions",
type=int,
default=10,
help=
"Only print n most common confusions. Defaults to 10, use -1 for all.")
parser.add_argument(
"--n_worst_lines",
type=int,
default=0,
help="Print the n worst recognized text lines with its error")
parser.add_argument(
"--xlsx_output",
type=str,
help="Optionally write a xlsx file with the evaluation results")
parser.add_argument("--num_threads",
type=int,
default=1,
help="Number of threads to use for evaluation")
parser.add_argument(
"--non_existing_file_handling_mode",
type=str,
default="error",
help=
"How to handle non existing .pred.txt files. Possible modes: skip, empty, error. "
"'Skip' will simply skip the evaluation of that file (not counting it to errors). "
"'Empty' will handle this file as would it be empty (fully checking for errors)."
"'Error' will throw an exception if a file is not existing. This is the default behaviour."
)
parser.add_argument("--skip_empty_gt",
action="store_true",
default=False,
help="Ignore lines of the gt that are empty.")
parser.add_argument("--no_progress_bars",
action="store_true",
help="Do not show any progress bars")
parser.add_argument(
"--checkpoint",
type=str,
default=None,
help=
"Specify an optional checkpoint to parse the text preprocessor (for the gt txt files)"
)
# page xml specific args
parser.add_argument("--pagexml_gt_text_index", default=0)
parser.add_argument("--pagexml_pred_text_index", default=1)
args = parser.parse_args()
# check if loading a json file
if len(args.gt) == 1 and args.gt[0].endswith("json"):
with open(args.gt[0], 'r') as f:
json_args = json.load(f)
for key, value in json_args.items():
setattr(args, key, value)
print("Resolving files")
gt_files = sorted(glob_all(args.gt))
if args.pred:
pred_files = sorted(glob_all(args.pred))
else:
pred_files = [split_all_ext(gt)[0] + args.pred_ext for gt in gt_files]
args.pred_dataset = args.dataset
if args.non_existing_file_handling_mode.lower() == "skip":
non_existing_pred = [p for p in pred_files if not os.path.exists(p)]
for f in non_existing_pred:
idx = pred_files.index(f)
del pred_files[idx]
del gt_files[idx]
text_preproc = None
if args.checkpoint:
with open(
args.checkpoint if args.checkpoint.endswith(".json") else
args.checkpoint + '.json', 'r') as f:
checkpoint_params = json_format.Parse(f.read(), CheckpointParams())
text_preproc = text_processor_from_proto(
checkpoint_params.model.text_preprocessor)
non_existing_as_empty = args.non_existing_file_handling_mode.lower(
) != "error "
gt_data_set = create_dataset(
args.dataset,
DataSetMode.EVAL,
texts=gt_files,
non_existing_as_empty=non_existing_as_empty,
args={'text_index': args.pagexml_gt_text_index},
)
pred_data_set = create_dataset(
args.pred_dataset,
DataSetMode.EVAL,
texts=pred_files,
non_existing_as_empty=non_existing_as_empty,
args={'text_index': args.pagexml_pred_text_index},
)
evaluator = Evaluator(text_preprocessor=text_preproc,
skip_empty_gt=args.skip_empty_gt)
r = evaluator.run(gt_dataset=gt_data_set,
pred_dataset=pred_data_set,
processes=args.num_threads,
progress_bar=not args.no_progress_bars)
# TODO: More output
print("Evaluation result")
print("=================")
print("")
print(
"Got mean normalized label error rate of {:.2%} ({} errs, {} total chars, {} sync errs)"
.format(r["avg_ler"], r["total_char_errs"], r["total_chars"],
r["total_sync_errs"]))
# sort descending
print_confusions(r, args.n_confusions)
print_worst_lines(r, gt_data_set.samples(), args.n_worst_lines)
if args.xlsx_output:
write_xlsx(args.xlsx_output, [{
"prefix": "evaluation",
"results": r,
"gt_files": gt_files,
}])
def run(args):
# check if loading a json file
if len(args.files) == 1 and args.files[0].endswith("json"):
import json
with open(args.files[0], 'r') as f:
json_args = json.load(f)
for key, value in json_args.items():
if key == 'dataset' or key == 'validation_dataset':
setattr(args, key, DataSetType.from_string(value))
else:
setattr(args, key, value)
# parse whitelist
whitelist = args.whitelist
if len(whitelist) == 1:
whitelist = list(whitelist[0])
whitelist_files = glob_all(args.whitelist_files)
for f in whitelist_files:
with open(f) as txt:
whitelist += list(txt.read())
if args.gt_extension is None:
args.gt_extension = DataSetType.gt_extension(args.dataset)
if args.validation_extension is None:
args.validation_extension = DataSetType.gt_extension(args.validation_dataset)
if args.text_generator_params is not None:
with open(args.text_generator_params, 'r') as f:
args.text_generator_params = json_format.Parse(f.read(), TextGeneratorParameters())
else:
args.text_generator_params = TextGeneratorParameters()
if args.line_generator_params is not None:
with open(args.line_generator_params, 'r') as f:
args.line_generator_params = json_format.Parse(f.read(), LineGeneratorParameters())
else:
args.line_generator_params = LineGeneratorParameters()
dataset_args = {
'line_generator_params': args.line_generator_params,
'text_generator_params': args.text_generator_params,
'pad': args.dataset_pad,
'text_index': args.pagexml_text_index,
}
# Training dataset
dataset = create_train_dataset(args, dataset_args)
# Validation dataset
if args.validation:
print("Resolving validation files")
validation_image_files = glob_all(args.validation)
if not args.validation_text_files:
val_txt_files = [split_all_ext(f)[0] + args.validation_extension for f in validation_image_files]
else:
val_txt_files = sorted(glob_all(args.validation_text_files))
validation_image_files, val_txt_files = keep_files_with_same_file_name(validation_image_files, val_txt_files)
for img, gt in zip(validation_image_files, val_txt_files):
if split_all_ext(os.path.basename(img))[0] != split_all_ext(os.path.basename(gt))[0]:
raise Exception("Expected identical basenames of validation file: {} and {}".format(img, gt))
if len(set(val_txt_files)) != len(val_txt_files):
raise Exception("Some validation images are occurring more than once in the data set.")
validation_dataset = create_dataset(
args.validation_dataset,
DataSetMode.TRAIN,
images=validation_image_files,
texts=val_txt_files,
skip_invalid=not args.no_skip_invalid_gt,
args=dataset_args,
)
print("Found {} files in the validation dataset".format(len(validation_dataset)))
else:
validation_dataset = None
params = CheckpointParams()
params.max_iters = args.max_iters
params.stats_size = args.stats_size
params.batch_size = args.batch_size
params.checkpoint_frequency = args.checkpoint_frequency if args.checkpoint_frequency >= 0 else args.early_stopping_frequency
params.output_dir = args.output_dir
params.output_model_prefix = args.output_model_prefix
params.display = args.display
params.skip_invalid_gt = not args.no_skip_invalid_gt
params.processes = args.num_threads
params.data_aug_retrain_on_original = not args.only_train_on_augmented
params.early_stopping_frequency = args.early_stopping_frequency
params.early_stopping_nbest = args.early_stopping_nbest
params.early_stopping_best_model_prefix = args.early_stopping_best_model_prefix
params.early_stopping_best_model_output_dir = \
args.early_stopping_best_model_output_dir if args.early_stopping_best_model_output_dir else args.output_dir
if args.data_preprocessing is None or len(args.data_preprocessing) == 0:
args.data_preprocessing = [DataPreprocessorParams.DEFAULT_NORMALIZER]
params.model.data_preprocessor.type = DataPreprocessorParams.MULTI_NORMALIZER
for preproc in args.data_preprocessing:
pp = params.model.data_preprocessor.children.add()
pp.type = DataPreprocessorParams.Type.Value(preproc) if isinstance(preproc, str) else preproc
pp.line_height = args.line_height
pp.pad = args.pad
# Text pre processing (reading)
params.model.text_preprocessor.type = TextProcessorParams.MULTI_NORMALIZER
default_text_normalizer_params(params.model.text_preprocessor.children.add(), default=args.text_normalization)
default_text_regularizer_params(params.model.text_preprocessor.children.add(), groups=args.text_regularization)
strip_processor_params = params.model.text_preprocessor.children.add()
strip_processor_params.type = TextProcessorParams.STRIP_NORMALIZER
# Text post processing (prediction)
params.model.text_postprocessor.type = TextProcessorParams.MULTI_NORMALIZER
default_text_normalizer_params(params.model.text_postprocessor.children.add(), default=args.text_normalization)
default_text_regularizer_params(params.model.text_postprocessor.children.add(), groups=args.text_regularization)
strip_processor_params = params.model.text_postprocessor.children.add()
strip_processor_params.type = TextProcessorParams.STRIP_NORMALIZER
if args.seed > 0:
params.model.network.backend.random_seed = args.seed
if args.bidi_dir:
# change bidirectional text direction if desired
bidi_dir_to_enum = {"rtl": TextProcessorParams.BIDI_RTL, "ltr": TextProcessorParams.BIDI_LTR,
"auto": TextProcessorParams.BIDI_AUTO}
bidi_processor_params = params.model.text_preprocessor.children.add()
bidi_processor_params.type = TextProcessorParams.BIDI_NORMALIZER
bidi_processor_params.bidi_direction = bidi_dir_to_enum[args.bidi_dir]
bidi_processor_params = params.model.text_postprocessor.children.add()
bidi_processor_params.type = TextProcessorParams.BIDI_NORMALIZER
bidi_processor_params.bidi_direction = TextProcessorParams.BIDI_AUTO
params.model.line_height = args.line_height
network_params_from_definition_string(args.network, params.model.network)
params.model.network.clipping_mode = NetworkParams.ClippingMode.Value("CLIP_" + args.gradient_clipping_mode.upper())
params.model.network.clipping_constant = args.gradient_clipping_const
params.model.network.backend.fuzzy_ctc_library_path = args.fuzzy_ctc_library_path
params.model.network.backend.num_inter_threads = args.num_inter_threads
params.model.network.backend.num_intra_threads = args.num_intra_threads
params.model.network.backend.shuffle_buffer_size = args.shuffle_buffer_size
# create the actual trainer
trainer = Trainer(params,
dataset,
validation_dataset=validation_dataset,
data_augmenter=SimpleDataAugmenter(),
n_augmentations=args.n_augmentations,
weights=args.weights,
codec_whitelist=whitelist,
keep_loaded_codec=args.keep_loaded_codec,
preload_training=not args.train_data_on_the_fly,
preload_validation=not args.validation_data_on_the_fly,
)
trainer.train(
auto_compute_codec=not args.no_auto_compute_codec,
progress_bar=not args.no_progress_bars
)
def run(args):
# check if loading a json file
if len(args.files) == 1 and args.files[0].endswith("json"):
import json
with open(args.files[0], 'r') as f:
json_args = json.load(f)
for key, value in json_args.items():
setattr(args, key, value)
# checks
if args.extended_prediction_data_format not in ["pred", "json"]:
raise Exception(
"Only 'pred' and 'json' are allowed extended prediction data formats"
)
# add json as extension, resolve wildcard, expand user, ... and remove .json again
args.checkpoint = [(cp if cp.endswith(".json") else cp + ".json")
for cp in args.checkpoint]
args.checkpoint = glob_all(args.checkpoint)
args.checkpoint = [cp[:-5] for cp in args.checkpoint]
# create voter
voter_params = VoterParams()
voter_params.type = VoterParams.Type.Value(args.voter.upper())
voter = voter_from_proto(voter_params)
# load files
input_image_files = glob_all(args.files)
if args.text_files:
args.text_files = glob_all(args.text_files)
# skip invalid files and remove them, there wont be predictions of invalid files
dataset = create_dataset(
args.dataset,
DataSetMode.PREDICT,
input_image_files,
args.text_files,
skip_invalid=True,
remove_invalid=True,
args={'text_index': args.pagexml_text_index},
)
print("Found {} files in the dataset".format(len(dataset)))
if len(dataset) == 0:
raise Exception(
"Empty dataset provided. Check your files argument (got {})!".
format(args.files))
# predict for all models
predictor = MultiPredictor(checkpoints=args.checkpoint,
batch_size=args.batch_size,
processes=args.processes)
do_prediction = predictor.predict_dataset(
dataset, progress_bar=not args.no_progress_bars)
avg_sentence_confidence = 0
n_predictions = 0
# output the voted results to the appropriate files
for result, sample in do_prediction:
n_predictions += 1
for i, p in enumerate(result):
p.prediction.id = "fold_{}".format(i)
# vote the results (if only one model is given, this will just return the sentences)
prediction = voter.vote_prediction_result(result)
prediction.id = "voted"
sentence = prediction.sentence
avg_sentence_confidence += prediction.avg_char_probability
if args.verbose:
lr = "\u202A\u202B"
print("{}: '{}{}{}'".format(sample['id'],
lr[get_base_level(sentence)], sentence,
"\u202C"))
output_dir = args.output_dir
dataset.store_text(sentence,
sample,
output_dir=output_dir,
extension=".pred.txt")
if args.extended_prediction_data:
ps = Predictions()
ps.line_path = sample[
'image_path'] if 'image_path' in sample else sample['id']
ps.predictions.extend([prediction] +
[r.prediction for r in result])
output_dir = output_dir if output_dir else os.path.dirname(
ps.line_path)
if not os.path.exists(output_dir):
os.mkdir(output_dir)
if args.extended_prediction_data_format == "pred":
with open(os.path.join(output_dir, sample['id'] + ".pred"),
'wb') as f:
f.write(ps.SerializeToString())
elif args.extended_prediction_data_format == "json":
with open(os.path.join(output_dir, sample['id'] + ".json"),
'w') as f:
# remove logits
for prediction in ps.predictions:
prediction.logits.rows = 0
prediction.logits.cols = 0
prediction.logits.data[:] = []
f.write(
MessageToJson(ps, including_default_value_fields=True))
else:
raise Exception("Unknown prediction format.")
print("Average sentence confidence: {:.2%}".format(
avg_sentence_confidence / n_predictions))
dataset.store()
print("All files written")
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--version',
action='version',
version='%(prog)s v' + __version__)
parser.add_argument(
"--files",
nargs="+",
help=
"List all image files that shall be processed. Ground truth fils with the same "
"base name but with '.gt.txt' as extension are required at the same location",
required=True)
parser.add_argument(
"--text_files",
nargs="+",
default=None,
help="Optional list of GT files if they are in other directory")
parser.add_argument(
"--gt_extension",
default=None,
help="Default extension of the gt files (expected to exist in same dir)"
)
parser.add_argument("--dataset",
type=DataSetType.from_string,
choices=list(DataSetType),
default=DataSetType.FILE)
parser.add_argument("--line_height",
type=int,
default=48,
help="The line height")
parser.add_argument("--pad",
type=int,
default=16,
help="Padding (left right) of the line")
parser.add_argument("--processes",
type=int,
default=1,
help="The number of threads to use for all operations")
parser.add_argument("--n_cols", type=int, default=1)
parser.add_argument("--n_rows", type=int, default=5)
parser.add_argument("--select", type=int, nargs="+", default=[])
# text normalization/regularization
parser.add_argument(
"--n_augmentations",
type=float,
default=0,
help=
"Amount of data augmentation per line (done before training). If this number is < 1 "
"the amount is relative.")
parser.add_argument("--text_regularization",
type=str,
nargs="+",
default=["extended"],
help="Text regularization to apply.")
parser.add_argument(
"--text_normalization",
type=str,
default="NFC",
help="Unicode text normalization to apply. Defaults to NFC")
parser.add_argument("--data_preprocessing",
nargs="+",
type=DataPreprocessorParams.Type.Value,
choices=DataPreprocessorParams.Type.values(),
default=[DataPreprocessorParams.DEFAULT_NORMALIZER])
args = parser.parse_args()
# Text/Data processing
if args.data_preprocessing is None or len(args.data_preprocessing) == 0:
args.data_preprocessing = [DataPreprocessorParams.DEFAULT_NORMALIZER]
data_preprocessor = DataPreprocessorParams()
data_preprocessor.type = DataPreprocessorParams.MULTI_NORMALIZER
for preproc in args.data_preprocessing:
pp = data_preprocessor.children.add()
pp.type = preproc
pp.line_height = args.line_height
pp.pad = args.pad
# Text pre processing (reading)
text_preprocessor = TextProcessorParams()
text_preprocessor.type = TextProcessorParams.MULTI_NORMALIZER
default_text_normalizer_params(text_preprocessor.children.add(),
default=args.text_normalization)
default_text_regularizer_params(text_preprocessor.children.add(),
groups=args.text_regularization)
strip_processor_params = text_preprocessor.children.add()
strip_processor_params.type = TextProcessorParams.STRIP_NORMALIZER
text_preprocessor = text_processor_from_proto(text_preprocessor)
data_preprocessor = data_processor_from_proto(data_preprocessor)
print("Resolving input files")
input_image_files = sorted(glob_all(args.files))
if not args.text_files:
if args.gt_extension:
gt_txt_files = [
split_all_ext(f)[0] + args.gt_extension
for f in input_image_files
]
else:
gt_txt_files = [None] * len(input_image_files)
else:
gt_txt_files = sorted(glob_all(args.text_files))
input_image_files, gt_txt_files = keep_files_with_same_file_name(
input_image_files, gt_txt_files)
for img, gt in zip(input_image_files, gt_txt_files):
if split_all_ext(os.path.basename(img))[0] != split_all_ext(
os.path.basename(gt))[0]:
raise Exception(
"Expected identical basenames of file: {} and {}".format(
img, gt))
if len(set(gt_txt_files)) != len(gt_txt_files):
raise Exception(
"Some image are occurring more than once in the data set.")
dataset = create_dataset(
args.dataset,
DataSetMode.TRAIN,
images=input_image_files,
texts=gt_txt_files,
non_existing_as_empty=True,
)
if len(args.select) == 0:
args.select = range(len(dataset.samples()))
dataset._samples = dataset.samples()
else:
dataset._samples = [dataset.samples()[i] for i in args.select]
samples = dataset.samples()
print("Found {} files in the dataset".format(len(dataset)))
with StreamingInputDataset(
dataset,
data_preprocessor,
text_preprocessor,
SimpleDataAugmenter(),
args.n_augmentations,
) as input_dataset:
f, ax = plt.subplots(args.n_rows, args.n_cols, sharey='all')
row, col = 0, 0
for i, (id, sample) in enumerate(
zip(args.select, input_dataset.generator(args.processes))):
line, text, params = sample
if args.n_cols == 1:
ax[row].imshow(line.transpose())
ax[row].set_title("ID: {}\n{}".format(id, text))
else:
ax[row, col].imshow(line.transpose())
ax[row, col].set_title("ID: {}\n{}".format(id, text))
row += 1
if row == args.n_rows:
row = 0
col += 1
if col == args.n_cols or i == len(samples) - 1:
plt.show()
f, ax = plt.subplots(args.n_rows, args.n_cols, sharey='all')
row, col = 0, 0
def run(args):
# check if loading a json file
if len(args.files) == 1 and args.files[0].endswith("json"):
import json
with open(args.files[0], 'r') as f:
json_args = json.load(f)
for key, value in json_args.items():
setattr(args, key, value)
# parse whitelist
whitelist = args.whitelist
if len(whitelist) == 1:
whitelist = list(whitelist[0])
whitelist_files = glob_all(args.whitelist_files)
for f in whitelist_files:
with open(f) as txt:
whitelist += list(txt.read())
if args.gt_extension is None:
args.gt_extension = DataSetType.gt_extension(args.dataset)
if args.validation_extension is None:
args.validation_extension = DataSetType.gt_extension(args.validation_dataset)
# Training dataset
print("Resolving input files")
input_image_files = sorted(glob_all(args.files))
if not args.text_files:
gt_txt_files = [split_all_ext(f)[0] + args.gt_extension for f in input_image_files]
else:
gt_txt_files = sorted(glob_all(args.text_files))
input_image_files, gt_txt_files = keep_files_with_same_file_name(input_image_files, gt_txt_files)
for img, gt in zip(input_image_files, gt_txt_files):
if split_all_ext(os.path.basename(img))[0] != split_all_ext(os.path.basename(gt))[0]:
raise Exception("Expected identical basenames of file: {} and {}".format(img, gt))
if len(set(gt_txt_files)) != len(gt_txt_files):
raise Exception("Some image are occurring more than once in the data set.")
dataset = create_dataset(
args.dataset,
DataSetMode.TRAIN,
images=input_image_files,
texts=gt_txt_files,
skip_invalid=not args.no_skip_invalid_gt
)
print("Found {} files in the dataset".format(len(dataset)))
# Validation dataset
if args.validation:
print("Resolving validation files")
validation_image_files = glob_all(args.validation)
if not args.validation_text_files:
val_txt_files = [split_all_ext(f)[0] + args.validation_extension for f in validation_image_files]
else:
val_txt_files = sorted(glob_all(args.validation_text_files))
validation_image_files, val_txt_files = keep_files_with_same_file_name(validation_image_files, val_txt_files)
for img, gt in zip(validation_image_files, val_txt_files):
if split_all_ext(os.path.basename(img))[0] != split_all_ext(os.path.basename(gt))[0]:
raise Exception("Expected identical basenames of validation file: {} and {}".format(img, gt))
if len(set(val_txt_files)) != len(val_txt_files):
raise Exception("Some validation images are occurring more than once in the data set.")
validation_dataset = create_dataset(
args.validation_dataset,
DataSetMode.TRAIN,
images=validation_image_files,
texts=val_txt_files,
skip_invalid=not args.no_skip_invalid_gt)
print("Found {} files in the validation dataset".format(len(validation_dataset)))
else:
validation_dataset = None
params = CheckpointParams()
params.max_iters = args.max_iters
params.stats_size = args.stats_size
params.batch_size = args.batch_size
params.checkpoint_frequency = args.checkpoint_frequency if args.checkpoint_frequency >= 0 else args.early_stopping_frequency
params.output_dir = args.output_dir
params.output_model_prefix = args.output_model_prefix
params.display = args.display
params.skip_invalid_gt = not args.no_skip_invalid_gt
params.processes = args.num_threads
params.data_aug_retrain_on_original = not args.only_train_on_augmented
params.early_stopping_frequency = args.early_stopping_frequency
params.early_stopping_nbest = args.early_stopping_nbest
params.early_stopping_best_model_prefix = args.early_stopping_best_model_prefix
params.early_stopping_best_model_output_dir = \
args.early_stopping_best_model_output_dir if args.early_stopping_best_model_output_dir else args.output_dir
params.model.data_preprocessor.type = DataPreprocessorParams.DEFAULT_NORMALIZER
params.model.data_preprocessor.line_height = args.line_height
params.model.data_preprocessor.pad = args.pad
# Text pre processing (reading)
params.model.text_preprocessor.type = TextProcessorParams.MULTI_NORMALIZER
default_text_normalizer_params(params.model.text_preprocessor.children.add(), default=args.text_normalization)
default_text_regularizer_params(params.model.text_preprocessor.children.add(), groups=args.text_regularization)
strip_processor_params = params.model.text_preprocessor.children.add()
strip_processor_params.type = TextProcessorParams.STRIP_NORMALIZER
# Text post processing (prediction)
params.model.text_postprocessor.type = TextProcessorParams.MULTI_NORMALIZER
default_text_normalizer_params(params.model.text_postprocessor.children.add(), default=args.text_normalization)
default_text_regularizer_params(params.model.text_postprocessor.children.add(), groups=args.text_regularization)
strip_processor_params = params.model.text_postprocessor.children.add()
strip_processor_params.type = TextProcessorParams.STRIP_NORMALIZER
if args.seed > 0:
params.model.network.backend.random_seed = args.seed
if args.bidi_dir:
# change bidirectional text direction if desired
bidi_dir_to_enum = {"rtl": TextProcessorParams.BIDI_RTL, "ltr": TextProcessorParams.BIDI_LTR,
"auto": TextProcessorParams.BIDI_AUTO}
bidi_processor_params = params.model.text_preprocessor.children.add()
bidi_processor_params.type = TextProcessorParams.BIDI_NORMALIZER
bidi_processor_params.bidi_direction = bidi_dir_to_enum[args.bidi_dir]
bidi_processor_params = params.model.text_postprocessor.children.add()
bidi_processor_params.type = TextProcessorParams.BIDI_NORMALIZER
bidi_processor_params.bidi_direction = TextProcessorParams.BIDI_AUTO
params.model.line_height = args.line_height
network_params_from_definition_string(args.network, params.model.network)
params.model.network.clipping_mode = NetworkParams.ClippingMode.Value("CLIP_" + args.gradient_clipping_mode.upper())
params.model.network.clipping_constant = args.gradient_clipping_const
params.model.network.backend.fuzzy_ctc_library_path = args.fuzzy_ctc_library_path
params.model.network.backend.num_inter_threads = args.num_inter_threads
params.model.network.backend.num_intra_threads = args.num_intra_threads
# create the actual trainer
trainer = Trainer(params,
dataset,
validation_dataset=validation_dataset,
data_augmenter=SimpleDataAugmenter(),
n_augmentations=args.n_augmentations,
weights=args.weights,
codec_whitelist=whitelist,
preload_training=not args.train_data_on_the_fly,
preload_validation=not args.validation_data_on_the_fly,
)
trainer.train(
auto_compute_codec=not args.no_auto_compute_codec,
progress_bar=not args.no_progress_bars
)
def main():
parser = ArgumentParser()
parser.add_argument("--dataset", type=DataSetType.from_string, choices=list(DataSetType), default=DataSetType.FILE)
parser.add_argument("--gt", nargs="+", required=True,
help="Ground truth files (.gt.txt extension)")
parser.add_argument("--pred", nargs="+", default=None,
help="Prediction files if provided. Else files with .pred.txt are expected at the same "
"location as the gt.")
parser.add_argument("--pred_dataset", type=DataSetType.from_string, choices=list(DataSetType), default=DataSetType.FILE)
parser.add_argument("--pred_ext", type=str, default=".pred.txt",
help="Extension of the predicted text files")
parser.add_argument("--n_confusions", type=int, default=10,
help="Only print n most common confusions. Defaults to 10, use -1 for all.")
parser.add_argument("--n_worst_lines", type=int, default=0,
help="Print the n worst recognized text lines with its error")
parser.add_argument("--xlsx_output", type=str,
help="Optionally write a xlsx file with the evaluation results")
parser.add_argument("--num_threads", type=int, default=1,
help="Number of threads to use for evaluation")
parser.add_argument("--non_existing_file_handling_mode", type=str, default="error",
help="How to handle non existing .pred.txt files. Possible modes: skip, empty, error. "
"'Skip' will simply skip the evaluation of that file (not counting it to errors). "
"'Empty' will handle this file as would it be empty (fully checking for errors)."
"'Error' will throw an exception if a file is not existing. This is the default behaviour.")
parser.add_argument("--no_progress_bars", action="store_true",
help="Do not show any progress bars")
parser.add_argument("--checkpoint", type=str, default=None,
help="Specify an optional checkpoint to parse the text preprocessor (for the gt txt files)")
# page xml specific args
parser.add_argument("--pagexml_gt_text_index", default=0)
parser.add_argument("--pagexml_pred_text_index", default=1)
args = parser.parse_args()
print("Resolving files")
gt_files = sorted(glob_all(args.gt))
if args.pred:
pred_files = sorted(glob_all(args.pred))
else:
pred_files = [split_all_ext(gt)[0] + args.pred_ext for gt in gt_files]
args.pred_dataset = args.dataset
if args.non_existing_file_handling_mode.lower() == "skip":
non_existing_pred = [p for p in pred_files if not os.path.exists(p)]
for f in non_existing_pred:
idx = pred_files.index(f)
del pred_files[idx]
del gt_files[idx]
text_preproc = None
if args.checkpoint:
with open(args.checkpoint if args.checkpoint.endswith(".json") else args.checkpoint + '.json', 'r') as f:
checkpoint_params = json_format.Parse(f.read(), CheckpointParams())
text_preproc = text_processor_from_proto(checkpoint_params.model.text_preprocessor)
non_existing_as_empty = args.non_existing_file_handling_mode.lower() != "error "
gt_data_set = create_dataset(
args.dataset,
DataSetMode.EVAL,
texts=gt_files,
non_existing_as_empty=non_existing_as_empty,
args={'text_index': args.pagexml_gt_text_index},
)
pred_data_set = create_dataset(
args.pred_dataset,
DataSetMode.EVAL,
texts=pred_files,
non_existing_as_empty=non_existing_as_empty,
args={'text_index': args.pagexml_pred_text_index},
)
evaluator = Evaluator(text_preprocessor=text_preproc)
r = evaluator.run(gt_dataset=gt_data_set, pred_dataset=pred_data_set, processes=args.num_threads,
progress_bar=not args.no_progress_bars)
# TODO: More output
print("Evaluation result")
print("=================")
print("")
print("Got mean normalized label error rate of {:.2%} ({} errs, {} total chars, {} sync errs)".format(
r["avg_ler"], r["total_char_errs"], r["total_chars"], r["total_sync_errs"]))
# sort descending
print_confusions(r, args.n_confusions)
print_worst_lines(r, gt_data_set.samples(), pred_data_set.text_samples(), args.n_worst_lines)
if args.xlsx_output:
write_xlsx(args.xlsx_output,
[{
"prefix": "evaluation",
"results": r,
"gt_files": gt_files,
"gts": gt_data_set.text_samples(),
"preds": pred_data_set.text_samples()
}])
def run(args):
# check if loading a json file
if len(args.files) == 1 and args.files[0].endswith("json"):
import json
with open(args.files[0], 'r') as f:
json_args = json.load(f)
for key, value in json_args.items():
setattr(args, key, value)
# checks
if args.extended_prediction_data_format not in ["pred", "json"]:
raise Exception("Only 'pred' and 'json' are allowed extended prediction data formats")
# add json as extension, resolve wildcard, expand user, ... and remove .json again
args.checkpoint = [(cp if cp.endswith(".json") else cp + ".json") for cp in args.checkpoint]
args.checkpoint = glob_all(args.checkpoint)
args.checkpoint = [cp[:-5] for cp in args.checkpoint]
# create voter
voter_params = VoterParams()
voter_params.type = VoterParams.Type.Value(args.voter.upper())
voter = voter_from_proto(voter_params)
# load files
input_image_files = glob_all(args.files)
if args.text_files:
args.text_files = glob_all(args.text_files)
# skip invalid files and remove them, there wont be predictions of invalid files
dataset = create_dataset(
args.dataset,
DataSetMode.PREDICT,
input_image_files,
args.text_files,
skip_invalid=True,
remove_invalid=True,
args={'text_index': args.pagexml_text_index},
)
print("Found {} files in the dataset".format(len(dataset)))
if len(dataset) == 0:
raise Exception("Empty dataset provided. Check your files argument (got {})!".format(args.files))
# predict for all models
predictor = MultiPredictor(checkpoints=args.checkpoint, batch_size=args.batch_size, processes=args.processes)
do_prediction = predictor.predict_dataset(dataset, progress_bar=not args.no_progress_bars)
avg_sentence_confidence = 0
n_predictions = 0
# output the voted results to the appropriate files
for result, sample in do_prediction:
n_predictions += 1
for i, p in enumerate(result):
p.prediction.id = "fold_{}".format(i)
# vote the results (if only one model is given, this will just return the sentences)
prediction = voter.vote_prediction_result(result)
prediction.id = "voted"
sentence = prediction.sentence
avg_sentence_confidence += prediction.avg_char_probability
if args.verbose:
lr = "\u202A\u202B"
print("{}: '{}{}{}'".format(sample['id'], lr[get_base_level(sentence)], sentence, "\u202C" ))
output_dir = args.output_dir
dataset.store_text(sentence, sample, output_dir=output_dir, extension=".pred.txt")
if args.extended_prediction_data:
ps = Predictions()
ps.line_path = sample['image_path'] if 'image_path' in sample else sample['id']
ps.predictions.extend([prediction] + [r.prediction for r in result])
output_dir = output_dir if output_dir else os.path.dirname(ps.line_path)
if not os.path.exists(output_dir):
os.mkdir(output_dir)
if args.extended_prediction_data_format == "pred":
with open(os.path.join(output_dir, sample['id'] + ".pred"), 'wb') as f:
f.write(ps.SerializeToString())
elif args.extended_prediction_data_format == "json":
with open(os.path.join(output_dir, sample['id'] + ".json"), 'w') as f:
# remove logits
for prediction in ps.predictions:
prediction.logits.rows = 0
prediction.logits.cols = 0
prediction.logits.data[:] = []
f.write(MessageToJson(ps, including_default_value_fields=True))
else:
raise Exception("Unknown prediction format.")
print("Average sentence confidence: {:.2%}".format(avg_sentence_confidence / n_predictions))
dataset.store()
print("All files written")
