Summary of Adaptive Deviation Learning For Visual Anomaly Detection with Data Contamination, by Anindya Sundar Das et al.
Adaptive Deviation Learning for Visual Anomaly Detection with Data Contamination
by Anindya Sundar Das, Guansong Pang, Monowar Bhuyan
First submitted to arxiv on: 14 Nov 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 proposes a novel approach to visual anomaly detection that addresses the issue of contaminated training data, which is common in real-world scenarios. The existing methods for anomaly detection typically rely on clean, unlabeled normal samples, but our approach employs deviation learning to compute anomaly scores while accounting for noise. By adjusting instance weights based on relative importance and utilizing a constrained optimization problem within the deviation learning framework, our method resolves the issue of data contamination. This results in a more stable and robust approach that surpasses competing techniques on benchmark datasets like MVTec and VisA. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper tries to find unusual images by looking at how they are different from normal pictures. They want to make sure this works even when the training data is not perfect, which often happens in real-life situations. Usually, these methods get worse if the data has some noise or errors. The new method uses something called deviation learning to figure out how weird each image is, and it makes sure that the normal images don’t look too weird by comparing them to a known pattern. For weird images, it makes them look even weirder compared to this pattern. They tested their method on some standard datasets and found that it worked better than other methods. |
Keywords
» Artificial intelligence » Anomaly detection » Optimization
