Create parsynth-ocr-200k.py

parsynth-ocr-200k.py

@@ -0,0 +1,79 @@
+import csv
+import os
+
+import datasets
+
+logger = datasets.logging.get_logger(__name__)
+
+_CITATION = """"""
+
+_DESCRIPTION = """ParsynthOCR-200k: A synthetic dataset for OCR. (A 200k samples preview)"""
+
+_DOWNLOAD_URLS = {
+    "train": "https://huggingface.co/datasets/hezarai/parsynth-ocr-200k/resolve/main/annotations_train.csv",
+    "test": "https://huggingface.co/datasets/hezarai/parsynth-ocr-200k/resolve/main/annotations_test.csv",
+    "data": "https://huggingface.co/datasets/hezarai/parsynth-ocr-200k/resolve/main/images.zip",
+}
+
+ZIP_IMAGES_DIR = "parsynth-ocr-200k"
+
+
+class ParsynthOCR200KConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super(ParsynthOCR200KConfig, self).__init__(**kwargs)
+
+
+class ParsynthOCR200K(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [
+        ParsynthOCR200KConfig(
+            name="Parsynth200K",
+            version=datasets.Version("1.0.0"),
+            description=_DESCRIPTION,
+        ),
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "image_path": datasets.Value("string"),
+                    "text": datasets.Value("string"),
+                }
+            ),
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        """
+        Return SplitGenerators.
+        """
+
+        train_path = dl_manager.download_and_extract(_DOWNLOAD_URLS["train"])
+        test_path = dl_manager.download_and_extract(_DOWNLOAD_URLS["test"])
+        archive_path = dl_manager.download(_DOWNLOAD_URLS["data"])
+        images_dir = dl_manager.extract(archive_path) if not dl_manager.is_streaming else ""
+        images_dir = os.path.join(images_dir, ZIP_IMAGES_DIR)
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN, gen_kwargs={"annotations_file": train_path, "images_dir": images_dir}
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST, gen_kwargs={"annotations_file": test_path, "images_dir": images_dir}
+            ),
+        ]
+
+    def _generate_examples(self, annotations_file, images_dir):
+        logger.info("⏳ Generating examples from = %s", annotations_file)
+
+        with open(annotations_file, encoding="utf-8") as csv_file:
+            csv_reader = csv.reader(csv_file, quotechar='"', skipinitialspace=True)
+
+            # Skip header
+            next(csv_reader, None)
+
+            for id_, row in enumerate(csv_reader):
+                filename, text = row
+                image_path = os.path.join(images_dir, filename)
+                yield id_, {"image_path": image_path, "text": text}