update data loading script
Browse files
food_vision_199_classes.py
CHANGED
@@ -243,8 +243,7 @@ class Food199(datasets.GeneratorBasedBuilder):
|
|
243 |
supervised_keys=("image", "label"),
|
244 |
homepage=_HOMEPAGE,
|
245 |
citation=_CITATION,
|
246 |
-
license=_LICENSE
|
247 |
-
task_templates=[ImageClassification(image_column="image", label_column="label")],
|
248 |
)
|
249 |
|
250 |
def _split_generators(self, dl_manager):
|
@@ -252,9 +251,10 @@ class Food199(datasets.GeneratorBasedBuilder):
|
|
252 |
This function returns the logic to split the dataset into different splits as well as labels.
|
253 |
"""
|
254 |
csv = dl_manager.download("annotations_with_links.csv")
|
255 |
-
df = pd.read_csv(csv)
|
256 |
print("Downloaded annotations.csv")
|
257 |
-
df_train_annotations = df[["image", "
|
|
|
258 |
# df_test_annotations = df[["filename", "label"]][df["split"] == "test"].to_dict(orient="records")
|
259 |
|
260 |
return [
|
@@ -277,6 +277,7 @@ class Food199(datasets.GeneratorBasedBuilder):
|
|
277 |
"""
|
278 |
for id_, row in enumerate(annotations):
|
279 |
row["image"] = str(row.pop("image"))
|
280 |
-
row["label"] = row.pop("
|
|
|
281 |
yield id_, row
|
282 |
|
|
|
243 |
supervised_keys=("image", "label"),
|
244 |
homepage=_HOMEPAGE,
|
245 |
citation=_CITATION,
|
246 |
+
license=_LICENSE
|
|
|
247 |
)
|
248 |
|
249 |
def _split_generators(self, dl_manager):
|
|
|
251 |
This function returns the logic to split the dataset into different splits as well as labels.
|
252 |
"""
|
253 |
csv = dl_manager.download("annotations_with_links.csv")
|
254 |
+
df = pd.read_csv(csv, low_memory=False)
|
255 |
print("Downloaded annotations.csv")
|
256 |
+
df_train_annotations = df[["image", "label"]][df["split"] == "train"].to_dict(orient="records")
|
257 |
+
print(df_train_annotations[:5])
|
258 |
# df_test_annotations = df[["filename", "label"]][df["split"] == "test"].to_dict(orient="records")
|
259 |
|
260 |
return [
|
|
|
277 |
"""
|
278 |
for id_, row in enumerate(annotations):
|
279 |
row["image"] = str(row.pop("image"))
|
280 |
+
row["label"] = row.pop("label")
|
281 |
+
print(id_, row)
|
282 |
yield id_, row
|
283 |
|