split argument of Places365() should accept "test" #8741
Description
🚀 The feature
split argument of Places365() accepts "train-standard", "train-challenge" and "val" as the doc says as shown below:
from torchvision import datasets
train_standard_data = datasets.Places365(
root="data",
split="train-standard" # Here
)
train_challenge_data = datasets.Places365(
root="data",
split="train-challenge" # Here
)
val_data = datasets.Places365(
root="data",
split="val" # Here
)
print(len(train_standard_data)) # 1803460
print(len(train_challenge_data )) # 8026628
print(len(val_data)) # 36500But split argument of Places365() doesn't accept "test" even though there is the test data(328,500) both for Places365-Standard and Places365-Challenge as shown below:
from torchvision import datasets
test_data = datasets.Places365(
root="data",
split="test", # Here
) # Error
print(len(test_data))ValueError: Unknown value 'test' for argument split. Valid values are {'train-standard', 'train-challenge', 'val'}.
Motivation, pitch
So split argument of Places365() should accept "test" as shown below:
from torchvision import datasets
test_data = datasets.Places365(
root="data",
split="test", # Here
)
print(len(test_data)) # 328500Alternatives
No response
Additional context
No response