Summary of Efficient Prompt Tuning Of Large Vision-language Model For Fine-grained Ship Classification, by Long Lan et al.
Efficient Prompt Tuning of Large Vision-Language Model for Fine-Grained Ship Classification
by Long Lan, Fengxiang Wang, Xiangtao Zheng, Zengmao Wang, Xinwang Liu
First submitted to arxiv on: 13 Mar 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 This paper explores how to improve fine-grained ship classification in remote sensing using pre-trained Vision-Language Models (VLMs). Traditional supervised methods are limited by the availability of labeled data, but VLMs can learn from few-shot or zero-shot scenarios. The researchers found that directly fine-tuning VLMs for RS-FGSC often leads to overfitting and poor generalization to unseen classes. To address this, they introduced a novel prompt tuning technique with a hierarchical, multi-granularity design that integrates remote sensing ship priors through bias terms. This approach enhanced the model’s ability to learn discriminative ship features and discern intricate backgrounds. The paper also presents a comprehensive dataset, FGSCM-52, which expands existing datasets with more extensive data and detailed annotations for less common ship classes. Experimental evaluations show that their proposed method outperforms current state-of-the-art techniques. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study helps computers better identify different types of ships in satellite images. It uses special AI models called Vision-Language Models to do this. These models are good at learning from small amounts of data, which is helpful because it’s hard to get lots of labeled pictures of all the different ship types. The problem with using these models directly is that they can become too good at identifying the types of ships they’ve seen before and not very good at identifying new ones. To fix this, the researchers came up with a new way to help the model learn what makes each type of ship unique. They also created a big dataset of labeled images to help other researchers improve their own models. |
Keywords
» Artificial intelligence » Classification » Few shot » Fine tuning » Generalization » Overfitting » Prompt » Supervised » Zero shot
