ruchangyao/FlashVSR_Plus_Dataset

FlashVSR+ Dataset

Project Page | Paper | GitHub | Models

Overview

We release the training dataset used in FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching.

FlashVSR+ is an efficient one-step diffusion framework for high-quality video super-resolution, enabling spatially untiled 4K inference on a single 24GB RTX 4090 GPU.

This dataset is released to support research on video super-resolution, video restoration, and diffusion-based generative models.

Dataset Structure

The dataset contains two types of training data:

• Video data: Video samples organized into approximately 140 compressed TAR archives (bucket_*.tar).
• Image data: Image samples organized into 38 compressed shards.

The dataset is distributed in sharded archives to facilitate storage and large-scale data processing.

Download

Download the dataset using the Hugging Face CLI:

pip install -U huggingface_hub
hf download ruchangyao/FlashVSR_Plus_Dataset \
    --repo-type dataset \
    --local-dir ./FlashVSR_Plus_Dataset

Alternatively, download individual archives from the Files and versions page.

Usage

The dataset is intended for training and research in:

• Video super-resolution and restoration
• Diffusion-based image and video generation
• High-resolution visual detail reconstruction
• Distribution matching distillation

For pretrained models and inference instructions, please refer to our GitHub repository.

Note: The public GitHub repository currently provides inference code. Training code is not included in this release.

Citation

If you use this dataset in your research, please cite our paper:

@misc{yao2026flashvsrefficientvideosuperresolution,
  title={FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching},
  author={Ruchang Yao and Shi Guo and Junhao Zhuang and Mingde Yao and Xin Cai and Jiahao Zhan and Bin Fu and Xinyuan Chen and Yu Qiao and Tianfan Xue},
  year={2026},
  eprint={2610.17552},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2610.17552}
}

License

The dataset is released under the Apache 2.0 License.