UROP Proceedings 2021-22

School of Engineering Department of Computer Science and Engineering 97 AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHU Boan / CPEG Course: UROP1100, Summer Fisheye images are often distorted and twisted, thus they often require distortion rectification. The generation-based method is one of the most popular methods because of its label-free training and onestage rectification. However, it has two main problems. The first one is that the decoder of the network is overburdened for rebuilding the structure and content of the images simultaneously. The second problem is that the naive skip-connection delivers the images directly which may lead to distortion diffusion. To solve these two problems, this article introduces a new method which contains a new network architecture. The new architecture has a parallel progressively complementary structure, which is used to separate content reconstruction and structure correction, thus reducing the burden of the decoder. In addition, the new architecture is embedded with correction layers in skip connection, aiming to pre-correct the images before transferring with the help of the appearance flows in different layers. As a result, the decoder can readily rebuild a credible result with the less distorted features. This new method shows high superiority according to different experimental results. Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: CHEN Siyu / CPEG Course: UROP1100, Fall UROP2100, Spring This paper presents a summary of the two main algorithms we used for detecting surrounding buildings or sites and their practical implementation process through python. Mobile applications need to obtain information about surrounding buildings and sites before pervasive positioning. How to determine the required buildings and sites are implemented in cloud server side. In this paper, we surveyed the related work in this field and concluded two algorithms for detecting surrounding buildings or sites with some input parameters. We also implemented them into practice python code to validate their correctness. We believe this paper can make a relatively comprehensive summary and make contribution to further implementation of the project.