Dataset Viewer

The dataset viewer is not available for this subset.

Cannot get the split names for the config 'default' of the dataset.

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 82, in _split_generators
                  raise ValueError(
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 65, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                         ^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Low-Light Restoration Dataset

Paired low-light / normal-light image patches for training and evaluation.

Contents of `low-light.tar`

low-light/
├── train/        30,000 paired patches  (60,000 files)
├── val/          1,000 paired patches   (2,000 files)
├── test/         1,000 input patches    (no ground truth)
└── dataset.py    reference PyTorch Dataset (Python 3.10+)

All images are lossless WebP (.webp).

File naming

Every image is named <id>-<role>.webp where role ∈ {in, gt}:

Split	Files
`train/`	`<id>-in.webp` paired with `<id>-gt.webp` (30,000 pairs)
`val/`	`<id>-in.webp` paired with `<id>-gt.webp` (1,000 pairs)
`test/`	`<id>-in.webp` only — no GT is provided

<id> is opaque; do not parse it. Pairing is by exact stem match.

Quick start

from pathlib import Path
from torch.utils.data import DataLoader
from dataset import (
    PairedLowLightDataset, TestLowLightDataset,
    PairedCompose, PairedRandomCrop, PairedRandomFlip, PairedToTensor,
)

root = Path("low-light")

train_tf = PairedCompose([
    PairedRandomCrop(256),
    PairedRandomFlip(p_h=0.5),
    PairedToTensor(),
])

train_set = PairedLowLightDataset(root / "train", transform=train_tf)
val_set   = PairedLowLightDataset(root / "val",   transform=None)
test_set  = TestLowLightDataset(root / "test",    transform=None)

train_loader = DataLoader(train_set, batch_size=16, shuffle=True, num_workers=4)
val_loader   = DataLoader(val_set,   batch_size=1,  shuffle=False, num_workers=2)
test_loader  = DataLoader(test_set,  batch_size=1,  shuffle=False, num_workers=2)

for x, y in train_loader:        # x, y are float32 CHW tensors in [0, 1]
    ...

for x, stem in test_loader:      # test yields (input, stem)
    pred = model(x)
    save_image(pred, f"submission/{stem[0]}-in.webp")

Dataset classes (in `dataset.py`)

`PairedLowLightDataset(root, transform=None)`

For train/ and val/. Returns (input_tensor, gt_tensor).

`TestLowLightDataset(root, transform=None)`

For test/. Returns (input_tensor, stem) where stem lets you save predictions under the original filename.

Transform contract

A transform may be one of:

None — images are converted to float32 CHW tensors in [0, 1].
A single-image torchvision-style callable fn(pil) -> tensor — applied independently to input and GT. Use only for deterministic ops (ToTensor, Normalize). Random single-image transforms will desync the pair.
A pair-aware callable fn(in_pil, gt_pil) -> (in_tensor, gt_tensor), marked by setting fn.paired = True. The callable owns randomness and must apply the same geometric augmentation to both images.

The provided PairedCompose, PairedRandomCrop, PairedRandomFlip, PairedToTensor building blocks already follow contract #3.

Submission format

For each <id>-in.webp in test/, produce a restored image and save it as <id>-in.webp (or .png if preferred). Keep the original stem.

Evaluation pairs each prediction against the private ground truth held by the organizers — do not attempt to obtain or infer test GTs.

Requirements

Python 3.10+
torch, torchvision, Pillow (with WebP support; built into modern Pillow)

Downloads last month: -

Total file size:

19.2 GB