Summary of Online Zero-shot Classification with Clip, by Qi Qian et al.
Online Zero-Shot Classification with CLIP
by Qi Qian, Juhua Hu
First submitted to arxiv on: 23 Aug 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 online zero-shot transfer scenario, where images arrive in a random order and are classified without storing their representations. The proposed framework leverages the distribution of target data to improve performance in real-world applications. To tackle this challenge, the authors develop two online strategies: online label learning to model the target data distribution, and online proxy learning to mitigate the modality gap between images and text. The combination of these predicted labels achieves an accuracy of 78.94% on ImageNet without accessing the entire dataset. Experiments on other 13 downstream tasks with different vision encoders show a more than 3% improvement on average, demonstrating the effectiveness of the proposal. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you’re trying to identify objects in pictures without looking at the picture again. This paper is about a new way to do this that works really well. The method uses information from all the pictures it has seen before to make better predictions. It’s like learning from experience. The results show that this approach can improve accuracy by more than 3% compared to other methods. This could be useful for things like identifying objects in self-driving cars or medical imaging. |
Keywords
» Artificial intelligence » Zero shot
