Add persian_license_plate.py

persian_license_plate.py

@@ -0,0 +1,94 @@
+import os
+import numpy as np
+from PIL import Image
+import csv
+
+import datasets
+
+logger = datasets.logging.get_logger(__name__)
+
+_CITATION = """"""
+
+_DESCRIPTION = """Persian Licensee plate dataset. Primarily taken from AmirKabir University Challenge. 
+Annotation are provided by the authors"""
+
+_DOWNLOAD_URLS = {
+    "train": "https://huggingface.co/datasets/hezarai/persian-license-plate-v1/resolve/main/persian_license_plate_train.csv",
+    "val": "https://huggingface.co/datasets/hezarai/persian-license-plate-v1/resolve/main/persian_license_plate_val.csv",
+    "test": "https://huggingface.co/datasets/hezarai/persian-license-plate-v1/resolve/main/persian_license_plate_test.csv",
+    'train_dataset': "https://huggingface.co/datasets/hezarai/persian-license-plate-v1/resolve/main/persian_license_plate_train.zip",
+    'val_dataset': "https://huggingface.co/datasets/hezarai/persian-license-plate-v1/resolve/main/persian_license_plate_val.zip",
+    'test_dataset': "https://huggingface.co/datasets/hezarai/persian-license-plate-v1/resolve/main/persian_license_plate_test.zip",
+}
+
+
+class PersianLicensePlateConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super(PersianLicensePlateConfig, self).__init__(**kwargs)
+
+
+class PersianLicensePlate(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [
+        PersianLicensePlateConfig(
+            name="PersianLicensePlate",
+            version=datasets.Version("1.0.0"),
+            description=_DESCRIPTION,
+        ),
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "label": datasets.Value("string"),
+                    "filename": datasets.Value("string"),
+                    "image": datasets.Sequence(datasets.Value("int32")),
+                }
+            ),
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        """
+        Return SplitGenerators.
+        """
+
+        train_path = dl_manager.download_and_extract(_DOWNLOAD_URLS["train"])
+        val_path = dl_manager.download_and_extract(_DOWNLOAD_URLS['val'])
+        test_path = dl_manager.download_and_extract(_DOWNLOAD_URLS["test"])
+
+        archive_path = dl_manager.download(_DOWNLOAD_URLS['train_dataset'])
+        train_extracted_path = dl_manager.extract(archive_path) if not dl_manager.is_streaming else None
+
+        archive_path = dl_manager.download(_DOWNLOAD_URLS['val_dataset'])
+        val_extracted_path = dl_manager.extract(archive_path) if not dl_manager.is_streaming else None
+
+        archive_path = dl_manager.download(_DOWNLOAD_URLS['test_dataset'])
+        test_extracted_path = dl_manager.extract(archive_path) if not dl_manager.is_streaming else None
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN, gen_kwargs={"filepath": train_path, "dataset_dir": train_extracted_path}
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST, gen_kwargs={"filepath": test_path, "dataset_dir": test_extracted_path}
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION, gen_kwargs={"filepath": val_path, "dataset_dir": val_extracted_path}
+            ),
+        ]
+
+    def _generate_examples(self, filepath, dataset_dir):
+        logger.info("⏳ Generating examples from = %s", filepath)
+
+        with open(filepath, 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):
+                label, filename = row
+                image_path = os.path.join(dataset_dir, filename)
+                yield id_, {"label": label, "image_path": image_path}