.png){kind=icon}
返回
Integrating knowledge-based sparse representation for image detection
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2020.09.089.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Image detection has been a primary step and widely adopted in diverse domains, such as autopilot, risk inspection, and robot vision. The main body of current models for image detection treat this task as the feature-based recognition of multiple independent objects. The adoption of semantic relationships, which can provide auxiliary information for joint recognition, is still in its early stage. Inspired by deep dictionary learning, this work proposes a novel framework named SEGNN, which aims at finding and using the sparse representation knowledge to improve the result of image detection. According to this design, both the efficiency and rationality of reasoning over semantic relationships are well captured. The framework includes an improved GGNN model. It can efficiently extract the semantic informations from the image, and map it to sparse vector. The framework also applies a novel SimRank method, to justify the rationality of the semantic reasoning. The performance of the framework is validated extensively on two datasets Visual Genome (VG) and COCO. The experimental results and case study have demonstrated the effectiveness of the proposed framework. (c) 2021 Published by Elsevier B.V.
Keyword:
Image detection
Deep reasoning
Knowledge graph
Gated graph neural network
Simrank
Dictionary learning
期刊
IF:
6.5
论文数:
2.5W
被引数:
6.5W
机构
引用论文
Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotations视觉基因组: 使用众包密集图像注释连接语言和视觉
OAAI
OAAI分享收藏Comparative study of layer by layer assembled multilayer films based on graphene oxide and reduced graphene oxide on flexible polyurethane foam: flame retardant and smoke suppression properties.png)
RSC Advances
RSC AdvancesIF0
PREAI
分享收藏Testing the concurrent validity of a naturalistic upper extremity reaching task测试自然主义上肢到达任务的同时有效性OAAI
OAAI分享收藏
查看更多内容