School of Engineering Department of Computer Science and Engineering 85 AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: CHEN Chenle / COMP Course: UROP 1100, Summer The limitations of GPS accuracy in indoor environments have prompted the exploration of alternative localization methods, leading to the emergence of wifi-based solutions. This project delves into the realm of wifi-based localization, specifically focusing on the utilization of the fingerprint method to enhance indoor positioning precision. By collecting and comparing Signal Strength (RSSI) data, the study aims to identify optimal matches for location determination. Through an investigation of wifi data matching techniques, the project seeks to improve the reliability and accuracy of indoor localization. The findings of this research contribute to advancing the field of indoor positioning systems and offer insights into optimizing wifi-based localization methodologies for enhanced user experiences. AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: CHEN Jhao Song / DSCT Course: UROP 1000, Summer It is well known that companies usually limit the return data per request for APIs for resource management and their pricing strategies. In order to extract POIs in large scale, a few attempts were made, including searching by recursion, searching through ordered square area and through other platforms than google map. Tsim Sha Tsui was experimented and returned 10275 POIs. AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: CHI Yankuan / COSC Course: UROP 1100, Fall UROP 2100, Spring UROP 3100, Summer 3D visual localization is one of the fundamental tasks in computer vision and has received great popularity in recent years. A large amount of methods are developed to deal with this task. One dominant type of method is feature matching, namely finding the correspondences between 2D image and 3D representation. This type of method has been widely researched for its high accuracy and possibility for real-world application. For structure-based feature matching, one class of methods of this type, choosing a good 3D scene representation is one of the crucial factors for the quality of the model. Thus, several excellent 3D scene representations are invented, from traditional point clouds and meshes to recent NeRF and 3D Gaussian. Different representations have different properties, as well as their capabilities for visual localization. In this project, I explored different 3D visual localization methods, and focused on different 3D scene representations’ performances.
RkJQdWJsaXNoZXIy NDk5Njg=