Harold328/OmniBench-99

Overview

OmniBench-99 benchmark is published in OmniCreator, containing 99 videos with varied contents (i.e., Environment, Human/Animal, and Object), designed to offer a comprehensive platform for evaluating generative video editing, focusing on both editing types and scenarios.

Paper Link

Project Page

Dataset Structure

Unlike previous benchmarks that evaluate only four editing types, OmniBench-99 expands the scope to include both editing types and scenarios. Specifically:

• Environment: Scenarios are developed for Background, Weather, and Time edits.
• Object: Scenarios are created for Addition, Removal, and Replacement edits.
• Human/Animal: Scenarios are designed for Appearance and Motion/Pose edits.

Citation

@article{chen2024omnicreator,
  title={OmniCreator: Self-Supervised Unified Generation with Universal Editing},
  author={Chen, Haodong and Wang, Lan and Yang, Harry and Lim, Ser-Nam},
  journal={arXiv preprint arXiv:2412.02114},
  year={2024}
}

Dataset Card Contact

Haodong Chen's email: haroldchen328@gmail.com