The research paper of Lê Minh Quân, a valedictorian from VNUHCM-University of Science, has been accepted at the AAAI 2024 Conference, a leading AI conference. Lê Minh Quân, an alumnus of the Talented Bachelor class, is a researcher at the VNUHCM-University of Science and has been sent to study for a PhD at Stony Brook University.
The name of the research paper is “Maskdiff: modelling mask distribution with diffusion probabilistic model for few-shot instance segmentation” by Minh-Quan Le, Tam V. Nguyen, Trung-Nghia Le, Thanh-Toan Do, Minh N. Do, and Minh-Triet Tran. The article is the result of Quân’s graduation thesis, which won the Eureka Second Prize in 2022. The topic was developed with guidance from teachers at the University of Dayton, the University of Illinois at Urbana-Champaign, Monash University, and VNUHCM-University of Science.
The article highlights Quân’s research on artificial intelligence and his potential for further research.
![📌](https://static.xx.fbcdn.net/images/emoji.php/v9/t5/1.5/16/1f4cc.png)
Abstract: Few-shot instance segmentation extends the few-shot learning paradigm to the instance segmentation task, which tries to segment instance objects from a query image with a few annotated examples of novel categories. Conventional approaches have attempted to address the task via prototype learning, known as point estimation. However, this mechanism is susceptible to noise and suffers from bias due to a significant scarcity of data. To overcome the disadvantages of the point estimation mechanism, we propose a novel approach, dubbed MaskDiff, which models the underlying conditional distribution of a binary mask, which is conditioned on an object region and K-shot information. Inspired by augmentation approaches that perturb data with Gaussian noise for populating low data density regions, we model the mask distribution with a diffusion probabilistic model. In addition, we propose to utilize classifier-free guided mask sampling to integrate category information into the binary mask generation process. Without bells and whistles, our proposed method consistently outperforms state-of-the-art methods on both base and novel classes of the COCO dataset while simultaneously being more stable than existing methods.
Leave a Reply