Summary of How to Train Your Vit For Ood Detection, by Maximilian Mueller and Matthias Hein
How to train your ViT for OOD Detection
by Maximilian Mueller, Matthias Hein
First submitted to arxiv on: 21 May 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 This paper investigates the performance of Vision Transformers (ViTs) as out-of-distribution detectors when fine-tuned from publicly available checkpoints. The authors analyze a large pool of ViT models to determine how pretraining and finetuning schemes impact their ability to detect anomalies on popular benchmarks. They find that pretraining has a strong influence on which finetuning method works well, and that certain training recipes are effective only for specific types of out-of-distribution data. The study identifies a best-practice training recipe for ViTs in this task. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how good Vision Transformers are at spotting weird images when they’re fine-tuned from existing models. The researchers looked at many different versions of these models to see what makes some better than others. They found that the way the models were trained before using them for anomaly detection is really important. Some ways of training make some models great at finding certain kinds of weird images, but not all. The study shows a good way to train Vision Transformers for this task. |
Keywords
» Artificial intelligence » Anomaly detection » Pretraining » Vit
