School of Engineering Department of Computer Science and Engineering 118 Generative AI Supervisor: CHEN Qifeng / CSE Student: ZHOU Yukai / COSC Course: UROP 2100, Fall 3D Geometry Estimation from 2D visuals has long been a challenging yet fundamental task for computer vision. Recent studies have proved that latent diffusion models, together with the U-Net and attention architectures, can achieve state-of-the-art results in the estimation of the depth and camera parameters from images or videos, besides their well-known generative abilities. The report delves into the detailed implementations behind such estimation models, as well as a generative diffusion model, which accepts camera pose representations as its conditions. Simplification of Trajectory Streams Supervisor: CHENG Siu Wing / CSE Student: SIN Tsz Yin / COGBM Course: UROP 1100, Spring Trajectory simplification is a well-studied problem in computational geometry and has significant use cases in various research areas. While there are not many streaming algorithms with quality guarantees, Cheng et al. proposed two algorithms for streaming δ-simplification and k-simplification, respectively. This UROP project studies the previous work by Cheng et al. by implementing the algorithms proposed, performing empirical testing, and discussing the results. In this project, we implemented a line simplification variant of the original δ-simplification in. We also tested it on the Beijing taxi dataset and compared results to previous algorithms found in the literature on trajectory simplification to help evaluate the performance of the algorithm in. Using Large Language Models (LLMs) for Software Development Supervisor: CHEUNG Shing Chi / CSE Student: CHAN Wai Pang / COMP Course: UROP 2100, Fall UROP 3100, Spring This semester, in collaboration with Prof. Cheung, Songqiang, and Yiuwa, we worked on a project to create a benchmark for evaluating the performance of large language models (LLMs) in code translation tasks. My primary contributions included running LiveCodeBench using local open-source LLMs to evaluate code translation accuracy, developing driver programs to enable our benchmark to automatically execute and evaluate translated code against LeetCode solutions. This report provides an overview of my accomplishments during the project, the methodologies I employed for code analysis and translation, and the collaborative efforts that contributed to achieving our project goals. Our work aims to facilitate future research in automated code translation by providing a reliable and reproducible evaluation framework.
RkJQdWJsaXNoZXIy NDk5Njg=