摘要
视觉关系检测或视觉关系识别,不仅需要识别出图像中的目标以及他们的位置,还要识别目标之间的相互关系,是计算机视觉领域非常具有挑战性的任务,也是深度理解图像的基础。得益于近年深度学习的蓬勃发展,视觉关系检测技术取得了显著进步。本文介绍了近年来基于深度学习的视觉关系检测的研究进展,从主要挑战、应用领域、公开数据集、算法模型、模型评估标准、模型效果这几方面进行对比分析,并展望了视觉关系检测未来的发展方向和前景。
Besides identifying the objects and their positions in the images, visual relationship detection/visual relationship recognition also requires the identification of the interactions between the objects. Although visual relationship detection is a challenging task in the field of computer vision, the recent development of deep learning and significant advances in the techniques of visual relationship detection have laid the foundation for deep understanding of the images. This paper mainly reviews the research progress of visual relationship detection based on deep learning in recent years, compares and analyzes the main challenges, application fields, open data sets, algorithm models, model evaluation criteria, and model effects, and investigates the future development for visual relationship detection.
出处
《科技创新导报》
2019年第27期145-150,共6页
Science and Technology Innovation Herald
关键词
视觉关系
深度学习
语义模块
视觉模块
Visual relationships
Deep learning
Semantic module
Visual module
