Summary of Widin: Wording Image For Domain-invariant Representation in Single-source Domain Generalization, by Jiawei Ma et al.
WIDIn: Wording Image for Domain-Invariant Representation in Single-Source Domain Generalization
by Jiawei Ma, Yulei Niu, Shiyuan Huang, Guangxing Han, Shih-Fu Chang
First submitted to arxiv on: 28 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
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 The paper presents a self-supervision framework called WIDIn, which aims to create domain-invariant representations for vision encoder models. Current methods can extend the model’s vision capabilities to diverse distributions without explicit training, but still struggle with complexity at inference time due to coarse-grained image descriptions. WIDIn disentangles discriminative visual representation by leveraging a single domain and fine-grained alignment. This approach is applicable to both pre-trained vision-language models like CLIP and separately trained uni-modal models like MoCo and BERT. Experimental results on three domain generalization datasets demonstrate the effectiveness of WIDIn. |
Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper tries to make a computer vision model work better with different types of data without needing special training. The problem is that the current approach can only understand the big picture of an image, not the small details. The new method, called WIDIn, helps the model focus on the important parts and ignore the parts that are specific to each type of data. This makes the model better at understanding images it hasn’t seen before. Tests show that this approach works well for three different types of datasets. |
Keywords
» Artificial intelligence » Alignment » Bert » Domain generalization » Encoder » Inference