Summary of Diverse Representation Embedding For Lifelong Person Re-identification, by Shiben Liu et al.
Diverse Representation Embedding for Lifelong Person Re-Identification
by Shiben Liu, Huijie Fan, Qiang Wang, Xiai Chen, Zhi Han, Yandong Tang
First submitted to arxiv on: 24 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 The Lifelong Person Re-Identification (LReID) task aims to continuously learn from successive data streams, matching individuals across multiple cameras. The key challenge is preserving old knowledge while incrementally learning new information, caused by task-level domain gaps and limited old task datasets. Existing methods based on CNN backbones are insufficient for exploring instance representations from different perspectives, limiting model performance. Unlike these methods, the proposed Diverse Representations Embedding (DRE) framework uses a pure transformer for LReID, preserving old knowledge while adapting to new information. The Adaptive Constraint Module (ACM) integrates and pushes away operations between multiple overlapping representations generated by the transformer-based backbone, obtaining rich and discriminative instance representations. Based on these processed diverse representations, Knowledge Update (KU) and Knowledge Preservation (KP) strategies are proposed for task-level layout, enhancing adaptive learning ability and preserving old knowledge. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about a new way to recognize people across different cameras over time. The problem is that we need to remember what we learned before while also learning new things. Existing methods don’t do this well. This paper proposes a new approach called Diverse Representations Embedding, which uses a special kind of artificial intelligence called transformers. These transformers help us learn more about each person from different perspectives, making it better at recognizing people over time. |
Keywords
» Artificial intelligence » Cnn » Embedding » Transformer
