Summary of Efficient and Long-tailed Generalization For Pre-trained Vision-language Model, by Jiang-xin Shi et al.
Efficient and Long-Tailed Generalization for Pre-trained Vision-Language Model
by Jiang-Xin Shi, Chi Zhang, Tong Wei, Yu-Feng Li
First submitted to arxiv on: 18 Jun 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 presents a novel framework called Candle that addresses the challenges of adapting pre-trained vision-language models like CLIP to downstream tasks. Specifically, Candle tackles long-tailed data distributions and emerging tasks with new classes containing no samples. The framework proposes compensating logit-adjusted loss to encourage large margins of prototypes and alleviate imbalance. It also leverages extracted features to obtain visual and textual prototypes for prediction, and introduces cross-modal attention to enrich multi-modal information. Additionally, Candle introduces virtual prototypes for new classes to make up for their lack of training images. The approach achieves state-of-the-art performance on 11 diverse datasets while reducing training time. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Candle is a way to make pre-trained models like CLIP work better in real-world situations. These models are great at guessing what’s happening in pictures and videos, but they can struggle when the data isn’t balanced or there are new classes that don’t have any examples. Candle helps by making sure the model is fair and doesn’t favor certain classes over others. It also uses information from both images and text to make better predictions. This approach is useful for many applications, such as object detection, image classification, and more. |
Keywords
» Artificial intelligence » Attention » Image classification » Multi modal » Object detection
