报告题目：Dissecting Person Re-identification from the Viewpoint of Viewpoint
报 告 人：郑良 博士
报告时间：2019年4月12日（周五）上午10：30
报告地点：安徽大学磬苑校区理工D楼318会议室
报告摘要：Person re-identification is a cross-camera matching problem. It aims to retrieve all the images containing the same person with the person-of-interest. Intuitively, several environmental factors are considered influential to the system, including viewpoint, pose, illumination, background, illumination, etc. However, there lack rigid investigations to reveal what their real impact is. I will talk about our latest insights of how environmental factors influence a person re-id system. We use the person viewpoint as the study subject. An obstacle to our research is that the existing datasets are fixed, i.e., we cannot precisely have datasets with desired viewpoint distributions. To solve it, we introduce a large-scale synthetic data engine, PersonX. Composed of hand-crafted 3D person models, the salient characteristic of this engine is “controllable”. That is, we are able to synthesize pedestrians by setting the visual variables to arbitrary values. With PersonX, we can quantitatively analyze the influence of viewpoint on re-id accuracy. Comprehensively, viewpoints are precisely customized from 0 to 360 degree, allowing us to investigate its effect on the training, query, and gallery sets. Our research provides useful insights for dataset building and practical usage, e.g., a person of a side view makes a better query.
欢迎各位老师、同学届时前往！
科学技术处
2019年4月10日