ViP-CNN: Visual Phrase Guided Convolutional Neural Network | |
Yikang Li; Wanli Ouyang; Xiaogang Wang; Xiaoou Tang | |
2017 | |
会议地点 | 美国 |
英文摘要 | As the intermediate level task connecting image cap- tioning and object detection, visual relationship detection started to catch researchers’ attention because of its de- scriptive power and clear structure. It detects the objects and captures their pair-wise interactions with a subject- predicate-object triplet, e.g. hperson-ride-horsei. In this paper, each visual relationship is considered as a phrase with three components. We formulate the visual relationship detection as three inter-connected recognition problems and propose a Visual Phrase guided Convolutional Neural Net- work (ViP-CNN) to address them simultaneously. In ViP- CNN, we present a Phrase-guided Message Passing Struc- ture (PMPS) to establish the connection among relationship components and help the model consider the three problems jointly. Corresponding non-maximum suppression method and model training strategy are also proposed. Experimen- tal results show that our ViP-CNN outperforms the state- of-art method both in speed and accuracy. We further pre- train ViP-CNN on our cleansed Visual Genome Relation- ship dataset, which is found to perform better than the pre- training on the ImageNet for this task. |
语种 | 英语 |
内容类型 | 会议论文 |
源URL | [http://ir.siat.ac.cn:8080/handle/172644/11767] |
专题 | 深圳先进技术研究院_集成所 |
作者单位 | 2017 |
推荐引用方式 GB/T 7714 | Yikang Li,Wanli Ouyang,Xiaogang Wang,et al. ViP-CNN: Visual Phrase Guided Convolutional Neural Network[C]. 见:. 美国. |
个性服务 |
查看访问统计 |
相关权益政策 |
暂无数据 |
收藏/分享 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论