CORNE-Val

Paper | Code

CORNE-Val is an effect-aware object-removal benchmark containing 219 held-out image pairs. It is constructed from reserved NHR-Edit shards with the same SAVP procedure used to build CORNE.

Data Structure

CORNE-Val/
├── shot/          # Source images containing the target object
├── bg/            # Ground-truth object-free backgrounds
├── mask_sam/      # Object-core masks
└── mask-check/    # Effect-aware masks covering the object and its effects

Files with the same relative name correspond to one evaluation sample.

Evaluation

The two mask types support complementary evaluation settings:

• mask_sam/ evaluates removal from an object-core input region.
• mask-check/ provides an effect-aware reference region that includes associated visual effects.

Please use the same sample set and preprocessing settings when comparing methods.

Provenance and License

CORNE-Val is derived from NHR-Edit, which is released under the Apache License 2.0. CORNE-Val, including the OSOR-produced masks and metadata, is released under the Apache License 2.0. Users should retain the original NHR-Edit attribution.

Citation

If you use CORNE-Val, please cite both OSOR and NHR-Edit:

@inproceedings{zhou2026osor,
  title     = {OSOR: One-Step Diffusion Inpainting for Effect-Aware Object Removal},
  author    = {Zhou, Qinming and Sun, Chenxi and Kong, Deyang and He, Junhao and Tang, Xiangheng and Yu, Peike and Wu, Haotian and Cao, Leilei and Zhang, Linfeng},
  booktitle = {European Conference on Computer Vision (ECCV)},
  year      = {2026},
  url       = {https://arxiv.org/abs/2606.28094}
}

@article{Layer2025NoHumansRequired,
  arxivId      = {2507.14119},
  author       = {Maksim Kuprashevich and Grigorii Alekseenko and Irina Tolstykh and Georgii Fedorov and Bulat Suleimanov and Vladimir Dokholyan and Aleksandr Gordeev},
  title        = {NoHumansRequired: Autonomous High-Quality Image Editing Triplet Mining},
  year         = {2025},
  eprint       = {2507.14119},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV},
  url          = {https://arxiv.org/abs/2507.14119},
  journal      = {arXiv preprint arXiv:2507.14119}
}