Summary of Self-supervised Vision Transformers For Writer Retrieval, by Tim Raven et al.
Self-Supervised Vision Transformers for Writer Retrieval
by Tim Raven, Arthur Matei, Gernot A. Fink
First submitted to arxiv on: 1 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: 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 The paper proposes a novel method for writer retrieval using Vision Transformers (ViTs), which has achieved state-of-the-art performance in various domains but not yet in writer retrieval. The traditional methods in this domain rely on handcrafted features or Convolutional Neural Networks (CNNs). This work presents a self-supervised ViT-based feature extractor that aggregates VLAD-encoded features to improve writer retrieval performance. The results show that extracting local foreground features outperforms using the class token, achieving new state-of-the-art performances on two historical document collections: Historical-WI (83.1% mAP) and HisIR19 (95.0% mAP). Moreover, the ViT feature extractor can be directly applied to modern datasets like CVL (98.6% mAP) without fine-tuning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper uses a special kind of AI called Vision Transformers to help find writers in old documents. Right now, this is not something that’s been done well before, and the usual way people do it is by using things that are already known about writing or Convolutional Neural Networks (CNNs). This new method finds local parts of writing that make it easier to identify who wrote something. The results show that this works really well for finding writers in old documents, especially when compared to the traditional methods. |
Keywords
» Artificial intelligence » Fine tuning » Self supervised » Token » Vit
