Knowledge Commons of Institute of Automation,CAS
Enhancing Visual Question Answering Using Dropout | |
Fang, Zhiwei1,2![]() ![]() ![]() | |
2018-08 | |
会议名称 | ACM international conference on Multimedia |
会议日期 | 2018.10 |
会议地点 | Seoul, South Korea |
摘要 | Using dropout in Visual Question Answering (VQA) is a common practice to prevent overfitting. However, in multi-path networks, the current way to use dropout may cause two problems: the coadaptations of neurons and the explosion of output variance. In this paper, we propose the coherent dropout and the siamese dropout to solve the two problems, respectively. Specifically, in coherent dropout, all relevant dropout layers in multiple paths are forced to work coherently to maximize the ability of preventing neuron co-adaptations. We show that the coherent dropout is simple in implementation but very effective to overcome overfitting. As for the explosion of output variance, we develop a siamese dropout mechanism to explicitly minimize the difference between the two output vectors produced from the same input data during training phase. Such mechanism can reduce the gap between training and inference phases and make the VQA model more robust. Extensive experiments are conducted to verify the effectiveness of coherent dropout and siamese dropout. And the results also show that our methods can bring additional improvements on the state-of-the-art VQA models. |
收录类别 | SCI |
语种 | 英语 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/23597 |
专题 | 紫东太初大模型研究中心_图像与视频分析 |
通讯作者 | Jing, Liu |
作者单位 | 1.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, 95 Zhongguancun East Rd, Beijing 100190, Peoples R China 2.University of Chinese Academy of Sciences, Beijing, China 3.Business Growth BU, JD.com, China |
第一作者单位 | 模式识别国家重点实验室 |
通讯作者单位 | 模式识别国家重点实验室 |
推荐引用方式 GB/T 7714 | Fang, Zhiwei,Jing, Liu,Qiao, Yanyuan,et al. Enhancing Visual Question Answering Using Dropout[C],2018. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
Enhancing Visual Que(1635KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
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