Summary of Practicaldg: Perturbation Distillation on Vision-language Models For Hybrid Domain Generalization, by Zining Chen et al.
PracticalDG: Perturbation Distillation on Vision-Language Models for Hybrid Domain Generalization
by Zining Chen, Weiqiu Wang, Zhicheng Zhao, Fei Su, Aidong Men, Hongying Meng
First submitted to arxiv on: 13 Apr 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel approach to domain generalization, Open Set Domain Generalization (OSDG), which aims to address unseen classes from target domains. The authors introduce Perturbation Distillation (PD) from three perspectives: Score, Class, and Instance (SCI), named SCI-PD. This method leverages vision-language models (VLMs) to improve the robustness of lightweight vision models. Moreover, a new benchmark called Hybrid Domain Generalization (HDG) is proposed, along with a novel metric H^2-CV, which comprehensively assesses the robustness of algorithms by constructing various splits. The authors demonstrate that their method outperforms state-of-the-art algorithms on multiple datasets, particularly in scenarios with data scarcity. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In this paper, researchers are working to make computers better at recognizing objects when they look different from what the computer has seen before. They’re using special models called vision-language models to help computers learn about new things. The authors have a new way of training these models that works well even when there isn’t much data to work with. They also created a new test to see how good their method is, and it did better than other methods in some cases. |
Keywords
* Artificial intelligence * Distillation * Domain generalization
