Update dataset structure and provide legacy format export instructions

README.md

@@ -108,9 +108,74 @@ Baseline for Random guessing is 0.3.
 English, Latin
 
 ## Dataset Structure
+The repository consists of the following:
 
 ```
-/dataset/
+├── data
+│   ├── train-00000-of-00009.parquet
+│   ├── ...
+│   └── train-00008-of-00009.parquet
+├── metadata
+│   ├── licenses.csv
+│   └── rarespecies-catalog.csv
+├── metadata.csv
+├── README.md
+└── visuals
+    ├── phyla_ToL_tree.html
+    ├── phyla_ToL_tree.pdf
+    └── phyla_ToL_tree.png
+```
+To load the dataset, use the `datasets` library.
+```python
+from datasets import load_dataset
+
+ds = load_dataset("imageomics/rare-species", split="train")
+```
+
+You will have:
+```python
+ds
+Dataset({
+    features: ['file_name', 'rarespecies_id', 'eol_content_id', 'eol_page_id', 'kingdom', 'phylum', 'class', 'order', 'family', 'genus', 'species', 'sciName', 'common'],
+    num_rows: 11983
+})
+```
+Where, for example, an entry `ds[0]` contains:
+```python
+{ 'file_name': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=400x400 at 0x14F5F54C9940>,
+  'rarespecies_id': '75fd91cb-2881-41cd-88e6-de451e8b60e2',
+  'eol_content_id': '12853737',
+  'eol_page_id': '449393',
+  'kingdom': 'Animalia',
+  'phylum': 'Mollusca',
+  'class': 'Bivalvia',
+  'order': 'Unionida',
+  'family': 'Unionidae',
+  'genus': 'Cyclonaias',
+  'species': 'tuberculata',
+  'sciName': 'Cyclonaias tuberculata',
+  'common': 'purple wartyback'
+}
+```
+For more information about Hugging Face datasets, see the [documentation](https://huggingface.co/docs/datasets/loading#hugging-face-hub).
+
+
+To export the data into a format matching the previously used dataset structure, you may do the following without cloning the repository:
+
+1) Create and activate a virtual environment.
+2) Install dependencies.
+```bash
+pip install -r \
+  https://huggingface.co/datasets/imageomics/rare-species/resolve/main/scripts/requirements.txt
+```
+3) Run the export script. You may customize the output directory by specifying the `--dataset_path` argument.
+```bash
+curl -s https://huggingface.co/datasets/imageomics/rare-species/resolve/main/scripts/export_rare_species.py \
+  | python3 - --dataset_path ./exported_dataset
+```
+This will create a directory structure like the following:
+```
+/exported_dataset/dataset/
     <kingdom-phylum-class-order-family-genus-species-1>/
         <eol_content_id_1>_<eol_page_id>_eol_full-size-copy.jpg
         <eol_content_id_2>_<eol_page_id>_eol_full-size-copy.jpg

scripts/export_rare_species.py

@@ -0,0 +1,58 @@
+import os
+import argparse
+import polars as pl
+from datasets import load_dataset, Features, Value
+from datasets.features import Image
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Export the Rare Species dataset into a legacy on-disk folder structure"
+    )
+    parser.add_argument(
+        "--dataset-path", default="dataset",
+        help="Directory under which to write the `dataset/...` hierarchy"
+    )
+    parser.add_argument(
+        "--revision", default="main",
+        help="Hugging Face dataset revision (branch, tag, or commit SHA)"
+    )
+    args = parser.parse_args()
+
+    # Read metadata.csv from the remote
+    csv_url = (
+        f"https://huggingface.co/datasets/imageomics/rare-species/"
+        f"resolve/main/metadata.csv?download=true"
+    )
+    print(f"Loading metadata from {csv_url}")
+    df_pl = pl.read_csv(csv_url)
+    rel_paths = df_pl["file_name"].to_list()
+
+    # Define schema: file_name as raw bytes (no PIL decode), others as strings
+    features = Features({
+        "file_name": Image(decode=False),
+        **{c: Value("string") for c in df_pl.columns if c != "file_name"}
+    })
+
+    # Load the Parquet-backed dataset
+    print(f"Loading dataset imageomics/rare-species @ {args.revision}")
+    ds = load_dataset(
+        "imageomics/rare-species",
+        split="train",
+        revision=args.revision,
+        features=features
+    )
+
+    # Export each image's raw bytes under <path>/dataset/...
+    print(f"Exporting {len(rel_paths)} images to {args.dataset_path}/")
+    for idx, rel in enumerate(rel_paths):
+        info = ds[idx]["file_name"]
+        img_bytes = info["bytes"]
+        dst = os.path.join(args.dataset_path, rel)
+        os.makedirs(os.path.dirname(dst), exist_ok=True)
+        with open(dst, "wb") as f:
+            f.write(img_bytes)
+
+    print(f"Export complete: images written under {args.dataset_path}/dataset")
+
+if __name__ == "__main__":
+    main()

scripts/requirements.txt

@@ -0,0 +1,4 @@
+datasets
+polars
+pandas
+huggingface_hub