refactor

README.md

@@ -42,8 +42,8 @@ dataset_info:
           '32': Tyttocharax
   splits:
   - name: train
-    num_bytes: 578234
+    num_bytes: 1068363405
     num_examples: 3068
-  download_size: 330476983
-  dataset_size: 578234
+  download_size: 330399200
+  dataset_size: 1068363405
 ---

amazonian_fish_classifier_data.py

@@ -14,9 +14,8 @@
 """TODO."""
 
 
-import os
-import pandas as pd
 import datasets
+from pathlib import Path
 
 _CITATION = """TODO"""
 
@@ -32,7 +31,6 @@ _LICENSE = "CC BY 4.0"
 
 _URLS = {
     "images": "https://smithsonian.figshare.com/ndownloader/files/31975544",
-    "labels": "https://smithsonian.figshare.com/ndownloader/files/31975646",
 }
 
 
@@ -94,23 +92,22 @@ class AmazonianFish(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
-        images = dl_manager.download_and_extract(_URLS["images"])
-        labels = dl_manager.download(_URLS["labels"])
-        df = pd.read_csv(labels)
-        labels = df.to_dict(orient="records")
+        images = dl_manager.download(_URLS["images"])
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
-                gen_kwargs={
-                    "images": os.path.join(images, "training_images"),
-                    "labels": labels,
-                },
+                gen_kwargs={"archive": dl_manager.iter_archive(images)},
             ),
         ]
 
-    def _generate_examples(self, images, labels):
-        for id_, example in enumerate(labels):
-            yield id_, {
-                "image": str(os.path.join(images, example["Genus"], example["Image_name"])),
-                "label": example["Genus"],
-            }
+    def _generate_examples(self, archive):
+        id_ = 0
+        for fname, fobject in archive:
+            if fname.startswith("._"):
+                continue
+            if Path(fname).suffix != ".jpg":
+                continue
+            image = fobject.read()
+            label = str(Path(fname).parts[-2])
+            id_ += 1
+            yield id_, {"image": image, "label": label}