release NER models

davarocr/__init__.py

@@ -16,6 +16,8 @@
 from .davar_layout import *
 from .davar_videotext import *
 from .davar_table import *
+from .davar_nlp_common import *
+from .davar_ner import *
 from .mmcv import *
 from .version import __version__
 

davarocr/davar_common/apis/inference.py

@@ -57,6 +57,9 @@ def init_model(config, checkpoint=None, device='cuda:0', cfg_options=None):
     elif cfg_types == "SPOTTER":
         from davarocr.davar_spotting.models.builder import build_spotter
         model = build_spotter(config.model, test_cfg=config.get('test_cfg'))
+    elif cfg_types == "NER":
+        from davarocr.davar_ner.models.builder import build_ner
+        model = build_ner(config.model, test_cfg=config.get('test_cfg'))
     else:
         raise NotImplementedError
 

davarocr/davar_common/core/__init__.py

@@ -8,13 +8,14 @@
 # Date           :    2021-05-20
 ##################################################################################################
 """
-from .builder import POSTPROCESS, build_postprocess
+from .builder import POSTPROCESS, build_postprocess, CONVERTERS, build_converter
 from .evaluation import DavarDistEvalHook, DavarEvalHook
 
 
 __all__ = ['POSTPROCESS',
            'build_postprocess',
-
+           'CONVERTERS',
+           'build_converter',
            "DavarEvalHook",
            "DavarDistEvalHook",
            ]

davarocr/davar_common/core/builder.py

@@ -8,19 +8,30 @@
 # Date           :    2020-05-31
 ##################################################################################################
 """
-from mmcv.utils import Registry
-from mmdet.models.builder import build
+from mmcv.utils import Registry, build_from_cfg
 
 POSTPROCESS = Registry('postprocess')
-
+CONVERTERS = Registry('converter')
 
 def build_postprocess(cfg):
     """ Build POSTPROCESS module
 
     Args:
-       cfg(dict): module configuration
+       cfg(mmcv.Config): module configuration
 
     Returns:
        obj: POSTPROCESS module
     """
-    return build(cfg, POSTPROCESS)
+    return build_from_cfg(cfg, POSTPROCESS)
+
+
+def build_converter(cfg):
+    """
+    Args:
+        cfg (mmcv.Config): model config):
+
+    Returns:
+        obj: CONVERTER module
+
+    """
+    return build_from_cfg(cfg, CONVERTERS)

davarocr/davar_ner/__init__.py

@@ -0,0 +1,14 @@
+"""
+##################################################################################################
+# Copyright Info :    Copyright (c) Davar Lab @ Hikvision Research Institute. All rights reserved.
+# Filename       :    __init__.py
+# Abstract       :
+
+# Current Version:    1.0.0
+# Date           :    2022-05-06
+##################################################################################################
+"""
+from .datasets import *
+from .models import *
+from .core import *
+

davarocr/davar_ner/core/__init__.py

@@ -0,0 +1,18 @@
+"""
+##################################################################################################
+# Copyright Info :    Copyright (c) Davar Lab @ Hikvision Research Institute. All rights reserved.
+# Filename       :    __init__.py
+# Abstract       :
+
+# Current Version:    1.0.0
+# Date           :    2022-05-06
+##################################################################################################
+"""
+from .evaluation import eval_ner_f1
+from .converters import SpanConverter, TransformersConverter
+
+__all__ = [
+    'eval_ner_f1',
+    'SpanConverter',
+    'TransformersConverter'
+]

davarocr/davar_ner/core/converters/__init__.py

@@ -0,0 +1,15 @@
+"""
+##################################################################################################
+# Copyright Info :    Copyright (c) Davar Lab @ Hikvision Research Institute. All rights reserved.
+# Filename       :    __init__.py
+# Abstract       :
+
+# Current Version:    1.0.0
+# Date           :    2022-05-06
+##################################################################################################
+"""
+from .transformers_converter import TransformersConverter
+from .span_converter import SpanConverter
+
+
+__all__ = ['TransformersConverter', 'SpanConverter']

davarocr/davar_ner/core/converters/base_converter.py

@@ -0,0 +1,50 @@
+"""
+##################################################################################################
+# Copyright Info :    Copyright (c) Davar Lab @ Hikvision Research Institute. All rights reserved.
+# Filename       :    base_converter.py
+# Abstract       :
+
+# Current Version:    1.0.0
+# Date           :    2022-05-06
+##################################################################################################
+"""
+from abc import ABCMeta, abstractmethod
+
+
+class BaseConverter(metaclass=ABCMeta):
+    """ Base converter, Convert between text, index and tensor for NER pipeline.
+    """
+    @abstractmethod
+    def convert_text2id(self, results):
+        """ Convert token to ids.
+
+        Args:
+            results (dict): A dict must containing the token key:
+                    - tokens (list]): Tokens list.
+        Returns:
+            dict: corresponding ids
+        """
+        pass
+
+    @abstractmethod
+    def convert_pred2entities(self, preds, masks, **kwargs):
+        """ Gets entities from preds.
+
+        Args:
+            preds (list): Sequence of preds.
+            masks (Tensor): The valid part is 1 and the invalid part is 0.
+        Returns:
+            list: List of entities.
+        """
+        pass
+
+    @abstractmethod
+    def convert_entity2label(self, labels):
+        """ Convert labeled entities to ids.
+
+        Args:
+            labels (list): eg:['B-PER', 'I-PER']
+        Returns:
+            dict: corresponding labels
+        """
+        pass

davarocr/davar_ner/core/converters/span_converter.py

@@ -0,0 +1,112 @@
+"""
+##################################################################################################
+# Copyright Info :    Copyright (c) Davar Lab @ Hikvision Research Institute. All rights reserved.
+# Filename       :    span_converter.py
+# Abstract       :
+
+# Current Version:    1.0.0
+# Date           :    2022-05-06
+##################################################################################################
+"""
+from seqeval.scheme import Tokens, IOBES
+from davarocr.davar_common.core import CONVERTERS
+from .transformers_converter import TransformersConverter
+
+
+@CONVERTERS.register_module()
+class SpanConverter(TransformersConverter):
+    """Span converter, converter for span model.
+    """
+    def _generate_labelid_dict(self):
+        label2id_dict = {label: i for i, label in enumerate(['O'] + self.label_list)}
+        id2label_dict = {value: key for key, value in label2id_dict.items()}
+        return label2id_dict, id2label_dict
+
+
+    def _extract_subjects(self, seq):
+        """Get entities from label sequence
+        """
+        entities = [(t.to_tuple()[1], t.to_tuple()[2], t.to_tuple()[3]) for t in Tokens(seq, IOBES).entities]
+        return entities
+
+
+    def convert_entity2label(self, labels):
+        """Convert labeled entities to ids.
+
+        Args:
+            labels (list): eg:['B-PER', 'I-PER']
+
+        Returns:
+            dict: corresponding ids
+        """
+        labels = self._labels_convert(labels, self.only_label_first_subword)
+        cls_token_at_end=self.cls_token_at_end
+        pad_on_left = self.pad_on_left
+        label2id = self.label2id_dict
+        subjects = self._extract_subjects(labels)#get entities
+        start_ids = [0] * len(labels)
+        end_ids = [0] * len(labels)
+        subjects_id = []
+        for subject in subjects:
+            label = subject[0]
+            start = subject[1]
+            end = subject[2]
+
+            #set label for span
+            start_ids[start] = label2id[label]
+            end_ids[end-1] = label2id[label]#the true position is end-1
+            subjects_id.append((label2id[label], start, end))
+
+        # Account for [CLS] and [SEP] with "- 2".
+        special_tokens_count = 2
+        if len(labels) > self.max_len - special_tokens_count:
+            start_ids = start_ids[: (self.max_len - special_tokens_count)]
+            end_ids = end_ids[: (self.max_len - special_tokens_count)]
+
+        #add sep
+        start_ids += [0]
+        end_ids += [0]
+        if cls_token_at_end:
+            #add [CLS] at end
+            start_ids += [0]
+            end_ids += [0]
+        else:
+            #add [CLS] at begin
+            start_ids = [0]+ start_ids
+            end_ids = [0]+ end_ids
+        padding_length = self.max_len - len(labels) - 2
+        if pad_on_left:
+            #pad on left
+            start_ids = ([0] * padding_length) + start_ids
+            end_ids = ([0] * padding_length) + end_ids
+        else:
+            #pad on right
+            start_ids += ([0] * padding_length)
+            end_ids += ([0] * padding_length)
+        res = dict(start_positions=start_ids, end_positions=end_ids)
+        return res
+
+    def convert_pred2entities(self, preds, masks, **kwargs):
+        """Gets entities from preds.
+
+        Args:
+            preds (list): Sequence of preds.
+            masks (tensor): The valid part is 1 and the invalid part is 0.
+        Returns:
+            list: List of [[[entity_type,
+                                entity_start, entity_end]]].
+        """
+        id2label = self.id2label
+        pred_entities = []
+        for pred in preds:
+            entities = []
+            entity = [0, 0, 0]
+            for tag in pred:
+                entity[0] = id2label[tag[0]]
+                entity[1] = tag[1] - 1
+                entity[2] = tag[2] - 1
+                entities.append(entity.copy())
+            pred_entities.append(entities.copy())
+        tokens_index = [index.cpu().numpy().tolist()[0] for index in kwargs['tokens_index']]
+        pred_entities = [self._labels_convert_ori(pred_entity, tokens_index) for pred_entity in pred_entities]
+        return pred_entities