Summary of Data Augmentation For Image Classification Using Generative Ai, by Fazle Rahat et al.
Data Augmentation for Image Classification using Generative AI
by Fazle Rahat, M Shifat Hossain, Md Rubel Ahmed, Sumit Kumar Jha, Rickard Ewetz
First submitted to arxiv on: 31 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); 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 Automated Generative Data Augmentation (AGA), a framework that leverages large language models (LLMs), diffusion models, and segmentation models to expand datasets. AGA combines object extraction using segment and superclass-based methods with prompt diversity through combinatorial complexity decomposition and affine subject manipulation. The approach preserves foreground authenticity while introducing background diversity. Evaluations on ImageNet, CUB, and iWildCam demonstrate 15.6% and 23.5% accuracy improvements for in- and out-of-distribution data compared to baseline models, respectively. Additionally, AGA achieves a 64.3% improvement in SIC score over baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making artificial intelligence (AI) models better by using more data. One way to do this is to create fake versions of real images and videos that are similar but not exactly the same. This helps AI learn from a larger dataset, which makes it perform better. The researchers developed a new method called AGA that uses big language models and other techniques to create these fake datasets while keeping the important parts (like objects) accurate and diverse. They tested their approach on several datasets and showed significant improvements in accuracy compared to previous methods. |
Keywords
» Artificial intelligence » Data augmentation » Prompt
