feat: script

makeup-detection-dataset.py

@@ -0,0 +1,93 @@
+import datasets
+import pandas as pd
+
+_CITATION = """\
+@InProceedings{huggingface:dataset,
+title = {makeup-detection-dataset},
+author = {TrainingDataPro},
+year = {2023}
+}
+"""
+
+_DESCRIPTION = """\
+The dataset consists of photos featuring the same individuals captured in two
+distinct scenarios - *with and without makeup*. The dataset contains a diverse
+range of individuals with various *ages, ethnicities and genders*. The images
+themselves would be of high quality, ensuring clarity and detail for each
+subject.
+In photos with makeup, it is applied **to only specific parts** of the face,
+such as *eyes, lips, or skin*.
+In photos without makeup, individuals have a bare face with no visible
+cosmetics or beauty enhancements. These images would provide a clear contrast
+to the makeup images, allowing for significant visual analysis.
+"""
+_NAME = 'makeup-detection-dataset'
+
+_HOMEPAGE = f"https://huggingface.co/datasets/TrainingDataPro/{_NAME}"
+
+_LICENSE = ""
+
+_DATA = f"https://huggingface.co/datasets/TrainingDataPro/{_NAME}/resolve/main/data/"
+
+
+class MakeupDetectionDataset(datasets.GeneratorBasedBuilder):
+    """Small sample of image-text pairs"""
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features({
+                'no_makeup': datasets.Image(),
+                'with_makeup': datasets.Image(),
+                'part': datasets.Value('string'),
+                'gender': datasets.Value('string'),
+                'age': datasets.Value('int8'),
+                'country': datasets.Value('string')
+            }),
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        no_makeup = dl_manager.download(f"{_DATA}no_makeup.tar.gz")
+        with_makeup = dl_manager.download(f"{_DATA}with_makeup.tar.gz")
+        annotations = dl_manager.download(f"{_DATA}{_NAME}.csv")
+        no_makeup = dl_manager.iter_archive(no_makeup)
+        with_makeup = dl_manager.iter_archive(with_makeup)
+        return [
+            datasets.SplitGenerator(name=datasets.Split.TRAIN,
+                                    gen_kwargs={
+                                        "no_makeup": no_makeup,
+                                        'with_makeup': with_makeup,
+                                        'annotations': annotations
+                                    }),
+        ]
+
+    def _generate_examples(self, no_makeup, with_makeup, annotations):
+        annotations_df = pd.read_csv(annotations, sep=';')
+
+        for idx, ((image_path, image),
+                  (mask_path, mask)) in enumerate(zip(no_makeup, with_makeup)):
+            yield idx, {
+                "no_makeup": {
+                    "path": image_path,
+                    "bytes": image.read()
+                },
+                "with_makeup": {
+                    "path": mask_path,
+                    "bytes": mask.read()
+                },
+                'part':
+                    annotations_df.loc[annotations_df['no_makeup'].str.lower() ==
+                                       image_path.lower()]['part'].values[0],
+                'gender':
+                    annotations_df.loc[annotations_df['no_makeup'].str.lower() ==
+                                       image_path.lower()]['gender'].values[0],
+                'age':
+                    annotations_df.loc[annotations_df['no_makeup'].str.lower() ==
+                                       image_path.lower()]['age'].values[0],
+                'country':
+                    annotations_df.loc[annotations_df['no_makeup'].str.lower() ==
+                                       image_path.lower()]['country'].values[0]
+            }