DetermiNet
收藏arXiv2023-09-07 更新2024-06-21 收录
下载链接：
https://github.com/clarence-lee-sheng/DetermiNet
下载链接
链接失效反馈官方服务：
资源简介：
DetermiNet是由新加坡科技设计大学和新加坡科技研究局前沿人工智能研究中心共同开发的第一个大规模诊断数据集，专注于确定词类。该数据集包含250,000个合成生成的图像和标题，基于25个确定词。数据集的主要任务是预测边界框以识别感兴趣的对象，受给定确定词的语义约束。DetermiNet旨在解决现有视觉接地模型在理解和量化任务上的局限性，特别是在处理复杂引用和量化任务时的不足。
DetermiNet is the first large-scale diagnostic dataset jointly developed by the Singapore University of Technology and Design and the Institute of Advanced Artificial Intelligence Research under the Agency for Science, Technology and Research (A*STAR), Singapore, focusing on determiner words. This dataset contains 250,000 synthetically generated images and their corresponding captions, which are based on 25 distinct determiners. The core task of this dataset is to predict bounding boxes to identify objects of interest, constrained by the semantic meaning of the given determiners. DetermiNet aims to address the limitations of existing visual grounding models in comprehension and quantification tasks, particularly their shortcomings when handling complex referring and quantification tasks.
提供机构：
新加坡科技设计大学1, 新加坡科技研究局前沿人工智能研究中心2
创建时间：
2023-09-07
