Institutional Repository of Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
Answer Distillation for Visual Question Answering | |
Fang, Zhiwei1,2; Liu, Jing1; Tang, Qu1; Li, Yong3; Lu, Hanqing1 | |
2019-05 | |
会议名称 | Asian Conference on Computer Vision |
会议日期 | 2018.12 |
会议地点 | Perth, Australia |
摘要 | Answering open-ended questions in Visual Question Answering (VQA) is a challenging task. As the answers are totally free-form, the answer space for open-ended questions is innite in theory. This increases the diffculty for algorithms to predict the correct answers. In this paper, we propose a method named answer distillation to decrease the scale of answer space and limit the correct result into a small set of answer candidates. Specically, we design a two-stage architecture to answer a question: First, we develop an answer distillation network to distill the answers, converting an open-ended question to a multiple-choice one with a short list of answer candidates. Then, we make full use of the knowledge from the answer candidates to guide the visual attention and rene the prediction results. Extensive experiments are conducted to validate the effiectiveness of our answer distillation architecture. The results show that our method can effiectively compress the answer space and improve the accuracy on open-ended task, providing a new state-of-the-art performance on COCO-VQA dataset. |
语种 | 英语 |
七大方向——子方向分类 | 多模态智能 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/23599 |
专题 | 模式识别国家重点实验室_图像与视频分析 |
通讯作者 | Liu, Jing |
作者单位 | 1.Institute of Automation, Chinese Academy of Sciences, Beijing, China 2.University of Chinese Academy of Sciences, Beijing, China 3.Business Growth BU, JD.com |
第一作者单位 | 中国科学院自动化研究所 |
通讯作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Fang, Zhiwei,Liu, Jing,Tang, Qu,et al. Answer Distillation for Visual Question Answering[C],2019. |
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 | ||
Answer Distillation (2077KB) | 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