Re-Anime600 is made available exclusively for non-commercial research. Please provide the information below and accept the Re-Anime600 Non-Commercial Research License v1.0 before requesting access.

By requesting access to Re-Anime600, you acknowledge that the dataset contains material originating from third-party sources and previous research works, and that the dataset maintainers do not claim ownership of all underlying copyrighted material. Access is granted solely for non-commercial research subject to the license and terms of use below.

Re-Anime600

Re-Anime600 is a curated benchmark dataset for anime video quality assessment and super-resolution evaluation.

Our goal is to support and promote reproducible research on Deep Learning for Animation, including video restoration, super-resolution, quality assessment, enhancement, and related problems in computational imaging and generative AI.

Re-Anime600 contains 600 curated anime video clips selected to support the evaluation and development of methods operating on animated visual content.

The dataset is released for non-commercial research purposes only.

AINIME — AI for Animation

Re-Anime600 is also connected to AINIME — AI for Animation, a workshop dedicated to research at the intersection of artificial intelligence and animation.

AINIME brings together research on animation generation, motion control, restoration, super-resolution, quality evaluation, post-production, and responsible creative AI.

Re-Anime600 serves as the benchmark for the Anime Restoration Challenge, focused on anime video restoration and upscaling.

Citation

If you use Re-Anime600 in your research, please cite:

@InProceedings{Fargetta_2026_CVPR,
    author    = {Fargetta, Georgia and Conde, Marcos V. and Tanke, Julian and Timofte, Radu and Battiato, Sebastiano},
    title     = {Re-Anime600: A Curated Benchmark Dataset for Anime Video Quality Assessment and Super-Resolution Evaluation},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month     = {June},
    year      = {2026},
    pages     = {9005--9010}
}

Challenge

The AINIME Anime Restoration Challenge @ WACV 2027 is built on Re-Anime600.

Challenge page: https://mv-lab.github.io/ainime/
Training data and submission example: https://huggingface.co/datasets/ainime-challenge/re-anime600-train, https://github.com/Zero9667/ainime-wacv2027-submission-example
Register your team: https://docs.google.com/forms/d/e/1FAIpQLSfnYjhFO2Me3EcyD2xPB0Y_fDNgX97gcSpjFrPR17UTIuKwjw/viewform
Submit your code: https://docs.google.com/forms/d/e/1FAIpQLSdoIT-48PqB8ONCRU6ypvAZxR6VIQ1RW3enSqbDBAlyxyPPLQ/viewform

Disclaimer

Re-Anime600 is provided for research purposes on an "AS IS" basis, without warranties or representations of any kind regarding accuracy, completeness, fitness for a particular purpose, non-infringement, or the legal status of individual items contained in the dataset.

The authors and maintainers make no representation that they own all underlying copyrighted material contained in Re-Anime600.

Users are solely responsible for their use of the dataset and for any datasets, models, media, outputs, or other derivative materials they create from it.

See the LICENSE file for the complete terms governing access and use.

Terms of use

Access to Re-Anime600 is provided under the Re-Anime600 Non-Commercial Research License v1.0, available in the LICENSE file.

By accessing, downloading, or using the dataset, you agree that:

Research use only. Re-Anime600 may be used only for research purposes.
Non-commercial use only. The dataset may not be used for commercial applications, commercial products, commercial services, or other uses primarily intended for commercial advantage.
Citation is required. Publications, technical reports, benchmarks, presentations, and other research outputs substantially using Re-Anime600 must cite the Re-Anime600 paper.
Derivatives must preserve these terms. Any derivative dataset that you distribute, to the extent that you have the legal right to distribute it, must include this license, preserve applicable copyright and attribution notices, and remain subject to the same non-commercial research restrictions.
Third-party rights are not granted. Access to Re-Anime600 does not grant rights in third-party copyrighted material. Users remain responsible for determining whether additional permissions are required for their particular use.
Users are responsible for derivatives and outputs. The Re-Anime600 authors and maintainers are not responsible for derivative datasets, models, generated outputs, harmful content, unlawful content, or other material created through the use or misuse of this dataset.

For the complete terms, see LICENSE.

Dataset provenance and copyright

Re-Anime600 is a curated research dataset assembled using material originating from previous research works and datasets.

The Re-Anime600 authors and maintainers do not claim ownership of all underlying images, video frames, characters, artwork, or other copyrighted material contained in the dataset. Rights in such material remain with their respective copyright holders.

The license accompanying Re-Anime600 applies only to rights that the dataset authors and maintainers are legally entitled to license, including, where applicable, the original curation, selection, organization, metadata, annotations, and other contributions made specifically for Re-Anime600.

Access to this dataset does not constitute a transfer or license of third-party copyrights or other intellectual-property rights. Users are responsible for ensuring that their use of individual materials complies with applicable law and with any rights or restrictions associated with the original sources.

Copyright removal requests

If you are a copyright holder or authorized representative and believe that specific material included in Re-Anime600 should be removed, please contact the repository maintainers and identify the relevant material.

We will review reasonable removal requests and, where appropriate, remove the identified material from future distributions of the dataset.

Downloads last month: 196

Total file size:

3.48 GB