The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.

Couldn't infer the same data file format for all splits. Got {NamedSplit('train'): ('webdataset', {}), NamedSplit('validation'): ('csv', {}), NamedSplit('test'): ('csv', {})}

Error code:   FileFormatMismatchBetweenSplitsError

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.

Open Images V7 (WebDataset 10GB Tar Archives)

This repository hosts the Open Images V7 dataset packaged into standalone ~10 GB .tar archives for high-throughput streaming, fast downloads, and deep learning training with PyTorch / WebDataset.

Dataset Structure

validation/: 41,620 images (~12.02 GB) in 2 tar archives + segmentation labels & metadata.
test/: 125,436 images (~36.28 GB) in 4 tar archives + segmentation labels & metadata.
train/: 1,743,042 images (~508.46 GB) in 61 tar archives (sharded by hex 0..f) + segmentation labels & metadata.
Total Images: 1,910,098 images (~556.76 GB)
Labels: segmentations.csv and metadata (classes.csv, hierarchy.json, image_ids.csv, segmentation_classes.csv) included.

File Organization

├── validation/
│   ├── archives/validation_images_part_000.tar to 001.tar
│   ├── labels/segmentations.csv
│   └── metadata/classes.csv, hierarchy.json, ...
├── test/
│   ├── archives/test_images_part_000.tar to 003.tar
│   ├── labels/segmentations.csv
│   └── metadata/classes.csv, hierarchy.json, ...
└── train/
    ├── archives/train_images_shard_{0..f}_part_{000..004}.tar (61 files)
    ├── labels/segmentations.csv
    └── metadata/classes.csv, hierarchy.json, ...

Quick Start with WebDataset & PyTorch

import webdataset as wds

# Stream validation split directly without downloading entire dataset
url = "https://huggingface.co/datasets/zAura-xec/openimagev7/resolve/main/validation/archives/validation_images_part_{000..001}.tar"
dataset = wds.WebDataset(url).decode("pil").to_tuple("jpg;png", "json")

for image, meta in dataset:
    # Process image stream
    break

Downloads last month: 219

Total file size:

599 GB