Summary of High Efficiency Image Compression For Large Visual-language Models, by Binzhe Li et al.
High Efficiency Image Compression for Large Visual-Language Models
by Binzhe Li, Shurun Wang, Shiqi Wang, Yan Ye
First submitted to arxiv on: 24 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Image and Video Processing (eess.IV)
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 proposed variable bitrate image compression framework consists of a pre-editing module and an end-to-end codec, optimized for large visual language models (LVLMs) using token-level distortion and rank. The framework achieves promising rate-accuracy performance on various LVLMs and tasks, outperforming the state-of-the-art coding standard, Versatile Video Coding. Additionally, experiments with multi-modal tasks demonstrate the robustness and generalization capability of the proposed framework. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper proposes a new way to compress images using large language models. Instead of trying to optimize the compression for a specific task or set of tasks, it uses the language model’s ability to understand and represent different types of data. This approach is designed to work well with various image datasets and tasks. The results show that this method performs better than current state-of-the-art methods in terms of both rate (how much space the compressed image takes up) and accuracy (how well the original image can be reconstructed). |
Keywords
» Artificial intelligence » Generalization » Language model » Multi modal » Token
