发布时间:2024-06-10 浏览次数:

告题目:Machine Learning in Compositional Generative AI


单位:Simon Fraser University




Large language models (LLMs) and diffusion models have captivated both practitioners and the public with their remarkable capabilities in generative AI, yet control over object-level generation and editing remains less explored. Moreover, the allure of deploying generative foundation models in self-driving and visual editing has promoted the need for deeper investigation into generative modeling. For this reason, our recent research focuses on designing an alternative generative model: compositional generative AI for content creation. Compositional modeling, a fundamental concept in both computer vision and computer graphics, involves creating visual scenes through the assembly of components, objects, or elements with precise placement and interaction. This approach not only boosts controllability for user-friendly editing but also enhances the performance of visual downstream tasks like object detection and semantic segmentation. In this talk, I will briefly introduce my recent progress in indoor scene synthesis and image composition and discuss future directions.


Hang Zhou was a Postdoctoral Researcher at Visual Computing Department, Simon Fraser University, Canada, working with Prof. Hao (Richard) Zhang from 2021-2023. Previously, he obtained a PhD degree from University of Science and Technology of China in 2020. His research is primarily focused on scene understanding, compositional modelling, shape analysis, image generation, and 3D multimedia security. He received Cyberspace Security Fellowship in 2018. He has won Chinese Academy of Sciences Outstanding Doctoral Dissertation Award in 2021. He has also won the best paper award at IJCAI Workshop on safety & security of deep learning in 2021.

邮编:230601 联系电话:0551-6290 1380
Copyright @ 2023 合肥工业大学计算机与信息学院 皖公网安备 34011102000080号 皖ICP备05018251号-1