Datasets:
metadata
license: mit
language:
- en
pretty_name: W-Bench
size: 10,000 instances (for evaluation)
description: >
The W-Bench is designed for the evaluation of image watermarking models,
containing photographs sourced from the COCO, Flickr, and ShareGPT4V datasets.