Summary of Mars: Paying More Attention to Visual Attributes For Text-based Person Search, by Alex Ergasti et al.
MARS: Paying more attention to visual attributes for text-based person search
by Alex Ergasti, Tomaso Fontanini, Claudio Ferrari, Massimo Bertozzi, Andrea Prati
First submitted to arxiv on: 5 Jul 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 presents a novel approach to text-based person search (TBPS), which is a challenging problem that requires learning representations that bridge text and image data within a shared latent space. The proposed MARS architecture enhances existing TBPS systems by introducing two key components: Visual Reconstruction Loss and Attribute Loss. The former encourages the model to learn more expressive representations and textual-visual relations, while the latter balances the contribution of different types of attributes in the person retrieval process. Experimental results on three datasets (CUHK-PEDES, ICFG-PEDES, and RSTPReid) demonstrate significant performance improvements over current state-of-the-art models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper solves a tricky problem called text-based person search. It’s hard because you have to match words with pictures of people. The new method is called MARS, which stands for Mae-Attribute-Relation-Sensitive. It’s better than previous methods because it helps the computer learn more about how words relate to pictures. This makes it better at finding the right pictures when you show it some words. The researchers tested their method on three sets of pictures and found that it worked much better than other methods. |
Keywords
» Artificial intelligence » Latent space » Mae
