REU Site: NSF Research Experience for Undergraduates in Computer Vision

这个项目代表了计算机视觉本科生研究经验的延续，该网站在过去的27年里一直在中佛罗里达大学成功运营。多年来，来自全国各地65所院校的约270名本科生参加了这一项目。每年，学生将参加为期12周的全日制暑期课程。在过去的几年里，相当一部分与会者能够准备一篇提交给裁判会议的论文，该论文被接受，然后参加会议提交论文。该计划将在参与者中培养以下技能：编写复杂的计算机程序的能力；就研究问题进行口头交流的能力；通过PowerPoint等技术展示自己的项目的能力；就项目做海报演示并回答问题的能力；思考现实生活中计算机视觉问题的能力；初步阅读计算机视觉研究论文的能力；对视觉问题的讨论做出贡献的能力；以及着手申请研究生院的能力。研究重点是开发新的算法来解决视觉地理定位和分割、目标检测、多目标跟踪、活动和事件识别等科学问题。它促进了理论和实践的发展；在对困难的视觉问题进行数学建模和分析以及开发算法方面，同时建立了在现实生活中演示这些解决方案的真实系统。该网站采用的教育模式包括由一个由一名教授和一名研究生组成的团队全天候指导；一个简化的短期课程，让参与者更早地开始他们的研究项目；每天与导师会面，计划全天的活动；在MatLab中进行培训，以迅速转变研究想法。参与者参加计算机视觉基础的短期课程，将自己与他们最渴望的项目主题相匹配，并花费足够的时间进行重点研究。一些可能的研究项目包括：位置感知语义分割；具有社会行为、模型、人类行为识别的多目标跟踪等。暑假结束后，学生可以选择全年跟踪，与教授一起就他们的项目写一份技术报告，准备GRE，并申请研究生课程。

This project represents a continuation of a Research Experience for Undergraduates site in Computer Vision which has operated successfully at the University of Central Florida for the past twenty-seven years. Approximately two hundred and seventy undergraduate students from sixty five institutions all over the country have participated in this program over the years. Each year, students will participate in a 12-week duration full-time Summer program. In past years, a substantial fraction of participants have been able to prepare a paper for submission to a refereed conference, have the paper accepted and then attend the conference to present the paper. The program will build these skills in participants: ability to write complex computer programs; ability to communicate orally about a research problem; ability to present one's project via technology such as PowerPoint; ability to do a poster presentation about the project, and answer questions; ability to think about computer vision issues in real life problems; ability to do initial reading of a research paper in computer vision; ability to contribute to discussion of a vision problem; and ability to embark on applying to graduate school. The research emphasizes development of new algorithms for solving scientific problems in visual geo-localization and segmentation, object detection, multi-target tracking, activity and event recognition. It advances both theory and practice; in mathematical modeling and analysis of difficult vision problems and developing algorithms, while at the same time building real systems for demonstrating those solutions in real life situations. The educational model employed by this site includes round-the-clock mentoring by a team that includes a professor and a graduate student; a streamlined short course that lets participants start their research projects sooner; daily meetings with mentors to plan activities throughout the day; training in MatLab for quick turnaround of research ideas. Participants take the short course on fundamentals of computer vision, match themselves to a project topic that they most desire, and spend sufficient time in focused research. Some possible research projects include: Location-Aware Semantic Segmentation; Multi-target Tracking with Social Behavior, Model, Human Action Recognition etc. After summer, students then can opt for follow-through over the year by working with the professors to write a technical report on their project, to prepare for the GREs and to apply to graduate programs.