Upload 2 files

cn_segment_dataset.py

@@ -0,0 +1,103 @@
+import pandas as pd
+from huggingface_hub import hf_hub_url
+import datasets
+import os
+
+_VERSION = datasets.Version("0.0.1")
+
+_DESCRIPTION = "TODO"
+_HOMEPAGE = "TODO"
+_LICENSE = "TODO"
+_CITATION = "TODO"
+
+_FEATURES = datasets.Features(
+    {
+        "image": datasets.Image(),
+        "segment": datasets.Image(),
+        "prompt": datasets.Value("string"),
+    },
+)
+
+METADATA_URL = hf_hub_url(
+    "vision-paper/cn_segment_dataset",
+    filename="train.jsonl",
+    repo_type="dataset",
+)
+
+IMAGE_URL = hf_hub_url(
+    "vision-paper/cn_segment_dataset",
+    filename="image.zip",
+    repo_type="dataset",
+)
+
+SEGMENT_URL = hf_hub_url(
+    "vision-paper/cn_segment_dataset",
+    filename="segment.zip",
+    repo_type="dataset",
+)
+
+_DEFAULT_CONFIG = datasets.BuilderConfig(name="default", version=_VERSION)
+
+
+class VTONHD_segmented_segment(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [_DEFAULT_CONFIG]
+    DEFAULT_CONFIG_NAME = "default"
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=_FEATURES,
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        metadata_path = dl_manager.download(METADATA_URL)
+        image_dir = dl_manager.download_and_extract(
+            IMAGE_URL
+        )
+        segment_dir = dl_manager.download_and_extract(
+            SEGMENT_URL
+        )
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                # These kwargs will be passed to _generate_examples
+                gen_kwargs={
+                    "metadata_path": metadata_path,
+                    "image_dir": image_dir,
+                    "segment_dir": segment_dir,
+                },
+            ),
+        ]
+
+    def _generate_examples(self, metadata_path, image_dir, segment_dir, reference_dir):
+        metadata = pd.read_json(metadata_path, lines=True)
+
+        for _, row in metadata.iterrows():
+            prompt = row["prompt"]
+
+            image_path = row["image"]
+            image_path = os.path.join(image_dir, image_path)
+            image = open(image_path, "rb").read()
+
+            segment_path = row["segment"]
+            segment_path = os.path.join(
+                segment_dir, row["segment"]
+            )
+            segment = open(segment_path, "rb").read()
+
+            yield row["image"], {
+                "prompt": prompt,
+                "image": {
+                    "path": image_path,
+                    "bytes": image,
+                },
+                "segment": {
+                    "path": segment_path,
+                    "bytes": segment,
+                },
+            }