| 基于语义连通图的场景图生成算法 |
| |
| 引用本文： | 姜有亮,张锋军,沈沛意,,张 亮,.基于语义连通图的场景图生成算法[J].南京师范大学学报,2022,0(2):048-55. |
| |
| 作者姓名： | 姜有亮 张锋军 沈沛意 张 亮 |
| |
| 作者单位： | (1.西安电子科技大学计算机科学与技术学院,陕西 西安 710071)(2.中国电子科技网络信息安全有限公司,四川 成都 610041)(3.西安电子科技大学西安市智能软件工程重点实验室,陕西 西安 710071) |
| |
| 摘 要： | 提出了基于语义连通图的场景图生成算法. 将关系检测过程分为关系建议和关系推理两步; 以目标检测算法得到的候选对象为节点集合,构建一个全连接图; 使用物体的类别信息和相对空间关系计算物体之间存在关系的概率; 通过设置阈值来删除图中的无效连接,得到稀疏的语义连通图; 使用图神经网络聚合物体节点的特征进行聚合,融合上下文信息. 根据语义连通图的连接关系,结合更新后的主语和宾语特征以及两个物体联合区域的特征,构建关系特征,预测图中的每条边对应的关系类别.
|
| 关 键 词： | 场景图生成 图卷积神经网络 目标检测 视觉关系检测 场景语义理解 |
Scene Graph Generation Based on Semantic Connected Graph |
| |
| Jiang Youliang,Zhang Fengjun,Shen Peiyi,' target="_blank" rel="external">,Zhang Liang,' target="_blank" rel="external">.Scene Graph Generation Based on Semantic Connected Graph[J].Journal of Nanjing Nor Univ: Eng and Technol,2022,0(2):048-55. |
| |
| Authors: | Jiang Youliang Zhang Fengjun Shen Peiyi " target="_blank">' target="_blank" rel="external"> Zhang Liang " target="_blank">' target="_blank" rel="external"> |
| |
| Affiliation: | (1.School of Computer Science and Technology,Xidian University,Xi’an 710071,China)(2.China Electronics Technology Cyber Security Co.,Ltd.,Chengdu 610041,China)(3.Xi’an Key Laboratory of Intelligent Software Engineering,Xidian University,Xi’an 710071,China) |
| |
| Abstract: | A scene graph generation algorithm based on semantic connected graph is proposed. Relationship detection process can be divided into two steps:relationship advice and reasoning. The detected object candidates are used as nodes to build one fully connected diagram. Object category and relative space information are used to calculate the relationship probability between objects. A threshold is utilized to remove the invalid connection and build the sparse semantic connected graph. A graph neural network method is used to aggregate the node feature representation with contextual information. At last,the relation category corresponding to each edge of the graph is classified according to the connectivity of the semantic connectivity graph by combining the updated feature representations of the subject and object,and the characteristics of the joint region of the two objects. |
| |
| Keywords: | scene graph generation graph convolution network object detection visual relationship detection scene semantic understanding |
|
| 点击此处可从《南京师范大学学报》浏览原始摘要信息 |
|
点击此处可从《南京师范大学学报》下载全文 |
|
