Generative models show great potential in generating new samples, which have been extensively investigated in 2D/3D vision tasks. Among them, the adversarial training based models, e.g. Generative Adversarial Nets (GAN), are proven more effective in sample quality, compared with those likelihood based models, e.g. Variational Auto-encoders (VAE), Normalizing Flow (NF), etc. However, GANs show limitation in density modelling or stable -training, making it unsuitable for those likelihood-based tasks, e.g. out-of-distribution detection. Recently, score-based diffusion models are studied, and related research explains its superiority in both sample quality and density modelling. In this task, we will explain the basic idea of score-based diffusion models, and explore its potential in 2D/3D vision tasks.
Short Bio
Jing Zhang is currently a Lecturer with School of Computing, the Australian National University (Canberra, Australia). Her main research interests are generative models, uncertainty estimation, weakly supervised learning. She won the Best Student Paper Prize at DICTA 2017, the Best Deep/Machine Learning Paper Prize at APSIPA ASC 2017 and the Best Paper Award Nominee at IEEE CVPR 2020.
More Details
- When: Wed 8 March 2023, at 1:00 pm (GMT+10)
- Speaker: Dr Jing Zhang (Australian National University)
- Host: Dr Xin Yu
- Zoom: https://uqz.zoom.us/j/82896549343
No.23-02 Seven Algorithms for the Same Task (Testing Uniformity)