Merge pull request #544 from raileymontalan/kvis_th_ocr

Closes #224 | Add/Update Dataloader KVIS Thai OCR

seacrowd/sea_datasets/kvis_th_ocr/kvis_th_ocr.py

@@ -0,0 +1,179 @@
+# coding=utf-8
+# Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from pathlib import Path
+from typing import Dict, List, Tuple
+
+import datasets
+
+from seacrowd.utils import schemas
+from seacrowd.utils.configs import SEACrowdConfig
+from seacrowd.utils.constants import Licenses, Tasks
+
+_CITATION = """\
+@INPROCEEDINGS{8584876,
+  author={Joseph, Ferdin Joe John and Anantaprayoon, Panatchakorn},
+  booktitle={2018 International Conference on Information Technology (InCIT)},
+  title={Offline Handwritten Thai Character Recognition Using Single Tier Classifier and Local Features},
+  year={2018},
+  volume={},
+  number={},
+  pages={1-4},
+  abstract={Handwritten character recognition is a conversion process of handwriting into machine-encoded text. Currently,
+  several techniques and methods are proposed to enhance accuracy of handwritten character recognition for many languages
+  spoken across the globe. In this project, a local feature-based approach is proposed to enhance the accuracy of handwritten
+  offline character recognition for Thai alphabets. In the experiment, through MATLAB, 100 images for each class of Thai
+  alphabets are collected and k-fold cross validation is applied to manage datasets to train and test. A gradient invariant
+  feature set consisting of LBP and shape features is extracted. The classification is operated by using query matching based
+  on Euclidean distance. The accuracy would be the percentage of correct classification for each class. For the result, the
+  highest accuracy is 68.96% which has 144-bit shape features and uniform pattern LBP for the features.},
+  keywords={Character recognition;Feature extraction;Shape;Genetic algorithms;Matlab;Gray-scale;Optical character recognition
+  software;Offline Character Recognition;Local Binary Pattern;Thai Handwriting},
+  doi={10.23919/INCIT.2018.8584876},
+  ISSN={},
+  month={Oct},
+  url={https://ieeexplore.ieee.org/document/8584876}}
+
+"""
+
+_DATASETNAME = "kvis_th_ocr"
+
+_DESCRIPTION = """\
+The KVIS Thai OCR Dataset contains scanned handwritten version of all 44 Thai characters obtained from 27 individuals. It
+consisted of 1079 images from 44 classes (letters). This dataset consists of all Thai consonants with different writing
+styles of various people from ages between 16 and 75. Vowels and intonation are not taken into consideration for the dataset
+collected.
+"""
+
+_HOMEPAGE = "https://data.mendeley.com/datasets/8nr3pbdk5c/1"
+
+_LANGUAGES = ["tha"]
+
+_LICENSE = Licenses.CC_BY_4_0.value
+
+_LOCAL = False
+
+_URLS = "https://prod-dcd-datasets-cache-zipfiles.s3.eu-west-1.amazonaws.com/8nr3pbdk5c-1.zip"
+
+_SUPPORTED_TASKS = [Tasks.OPTICAL_CHARACTER_RECOGNITION]
+
+_SOURCE_VERSION = "1.0.0"
+
+_SEACROWD_VERSION = "1.0.0"
+
+
+class KVISThaiOCRDataset(datasets.GeneratorBasedBuilder):
+    """
+    KVIS Thai OCR is a dataset for optical character recognition for Thai characters from https://data.mendeley.com/datasets/8nr3pbdk5c/1.
+    """
+
+    SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
+    SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
+    labels = ["ก", "ข", "ฃ", "ค", "ฅ", "ฆ", "ง", "จ", "ฉ", "ช", "ซ", "ฌ", "ญ", "ฎ", "ฏ", "ฐ", "ฑ", "ฒ", "ณ", "ด", "ต", "ถ", "ท", "ธ", "น", "บ", "ป", "ผ", "ฝ", "พ", "ฟ", "ภ", "ม", "ย", "ร", "ล", "ว", "ศ", "ษ", "ส", "ห", "ฬ", "อ", "ฮ"]
+
+    BUILDER_CONFIGS = [
+        SEACrowdConfig(
+            name=f"{_DATASETNAME}_source",
+            version=datasets.Version(_SOURCE_VERSION),
+            description=f"{_DATASETNAME} source schema",
+            schema="source",
+            subset_id=f"{_DATASETNAME}",
+        ),
+        SEACrowdConfig(
+            name=f"{_DATASETNAME}_seacrowd_imtext",
+            version=datasets.Version(_SOURCE_VERSION),
+            description=f"{_DATASETNAME} SEACrowd schema",
+            schema="seacrowd_imtext",
+            subset_id=f"{_DATASETNAME}",
+        ),
+    ]
+
+    def _info(self) -> datasets.DatasetInfo:
+        if self.config.schema == "source":
+            features = datasets.Features(
+                {
+                    "id": datasets.Value("string"),
+                    "file_path": datasets.Value("string"),
+                    "character": datasets.Value("string"),
+                }
+            )
+        elif self.config.schema == "seacrowd_imtext":
+            features = schemas.image_text_features(label_names=self.labels)
+        else:
+            raise ValueError(f"Invalid schema: {self.config.schema}")
+
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
+        """
+        Returns SplitGenerators.
+        """
+
+        dir = dl_manager.download_and_extract(_URLS)
+        path = dl_manager.extract(os.path.join(dir, "KVIS TOCR Dataset.zip"))
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "path": path,
+                    "split": "train",
+                },
+            ),
+        ]
+
+    def _generate_examples(self, path: Path, split: str) -> Tuple[int, Dict]:
+        """
+        Yields examples as (key, example) tuples.
+        """
+
+        idx = 0
+        path = list(os.walk(path))
+        for directory in path[1:]:
+            label = directory[0][-3:-2]
+            for file in directory[2]:
+                file_extension = str(file[-3:])
+                if file_extension == "jpg":
+                    file_id = str(file[:-4])
+                    file_path = os.path.join(directory[0], file)
+                    if self.config.schema == "source":
+                        data = {
+                            "id": file_id,
+                            "file_path": file_path,
+                            "character": label,
+                        }
+                        yield idx, data
+                        idx += 1
+                    elif self.config.schema == "seacrowd_imtext":
+                        data = {
+                            "id": file_id,
+                            "image_paths": [file_path],
+                            "texts": "",
+                            "metadata": {
+                                "context": "",
+                                "labels": [label],
+                            },
+                        }
+                        yield idx, data
+                        idx += 1
+                    else:
+                        raise ValueError(f"Invalid schema: {self.config.schema}")

seacrowd/utils/constants.py

@@ -128,6 +128,7 @@ class Tasks(Enum):
     OPTICAL_CHARACTER_RECOGNITION = "OCR"
     STYLIZED_IMAGE_CAPTIONING = "SIC"
     VISUALLY_GROUNDED_REASONING = "VGR"
+    OPTICAL_CHARACTER_RECOGNITION = "OCR"
 
     # VideoText
     VIDEO_CAPTIONING = "VC"
@@ -291,6 +292,7 @@ class Licenses(Enum):
     Tasks.FACT_CHECKING: None,
     Tasks.WORD_LIST: None,
     Tasks.MORALITY_CLASSIFICATION: "TEXT",
+    Tasks.OPTICAL_CHARACTER_RECOGNITION: "IMTEXT"
 }
 
 SCHEMA_TO_TASKS = defaultdict(set)