WAS-R, WAS-S
收藏资源简介:
WAS数据集由华中科技大学和Adobe公司联合创建,专注于艺术文本分割任务。WAS-R数据集包含7100张艺术文本图像,具有高质量的词级标注,包括四边形框、掩码和转录。WAS-S数据集则是一个合成数据集,利用大型多模态模型和扩散模型生成,旨在提高文本分割模型的准确性和泛化能力。这些数据集主要用于解决艺术文本在复杂场景中的分割问题,特别是在文本图像生成、编辑和风格转换等任务中。
The WAS dataset was jointly created by Huazhong University of Science and Technology and Adobe Inc., focusing on the artistic text segmentation task. The WAS-R dataset contains 7100 artistic text images with high-quality word-level annotations, including quadrilateral bounding boxes, segmentation masks and transcriptions. The WAS-S dataset is a synthetic dataset generated using large multimodal models and diffusion models, aiming to improve the accuracy and generalization ability of text segmentation models. These datasets are primarily used to address the segmentation problem of artistic text in complex scenarios, especially in tasks such as text image generation, editing and style transfer.
WAS: 艺术文本分割数据集及方法 (ECCV 2024)
数据集引用
在使用WAS数据集或本仓库时,请引用以下论文:
@article{xie2024was, title={WAS: Dataset and Methods for Artistic Text Segmentation}, author={Xie, Xudong and Li, Yuzhe and Liu, Yang and Zhang, Zhifei and Wang, Zhaowen and Xiong, Wei and Bai, Xiang}, booktitle={ECCV}, year={2024} }

- 1WAS: Dataset and Methods for Artistic Text Segmentation华中科技大学 · 2024年



