MaskOVEN-Wiki
收藏资源简介:
MaskOVEN-Wiki数据集是由北京理工大学和腾讯联合创建的,旨在支持像素级视觉实体链接任务。该数据集包含超过520万条标注,涵盖20,077个实体,主要来源于OVEN-Wiki数据集,并通过自动化的反向区域-实体标注框架生成。数据集的创建过程结合了知识增强和模型集成,确保了高精度的标注质量,标注成功率达到94.8%。该数据集主要应用于细粒度的视觉理解任务,如多模态知识图谱补全、视觉问答和图像检索,旨在解决复杂场景下的视觉实体链接问题。
MaskOVEN-Wiki dataset was jointly created by Beijing Institute of Technology and Tencent, aiming to support pixel-level visual entity linking tasks. This dataset contains over 5.2 million annotated instances, covering 20,077 entities, and is mainly derived from the OVEN-Wiki dataset, generated via an automated reverse region-entity annotation framework. The dataset construction process combines knowledge enhancement and model ensemble to ensure high-precision annotation quality, with an annotation success rate of 94.8%. It is mainly applied to fine-grained visual understanding tasks such as multimodal knowledge graph completion, visual question answering, and image retrieval, aiming to address visual entity linking issues in complex scenarios.
MaskOVEN-wiki
数据集概述
- 名称: MaskOVEN-wiki
- 全称: Pixel-Level Visual Entity Linking Dataset
- 类型: 像素级视觉实体链接数据集




