Summary of In or Out? Fixing Imagenet Out-of-distribution Detection Evaluation, by Julian Bitterwolf et al.
In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation
by Julian Bitterwolf, Maximilian Müller, Matthias Hein
First submitted to arxiv on: 1 Jun 2023
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
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 tackles the problem of out-of-distribution (OOD) detection, where models are evaluated on their ability to identify inputs that don’t belong to the original training data. The authors criticize existing OOD test datasets, which they claim have serious flaws, such as including objects from the same classes as the training data. They propose a new dataset called NINCO, designed to provide a more accurate evaluation of OOD detectors by ensuring each sample is truly out-of-distribution. The paper also introduces synthetic “OOD unit-tests” to further test the models’ strengths and weaknesses. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research is about finding ways for computers to detect when they’re given information that doesn’t belong to what they were trained on. This is important because it helps us understand how well our computer models can generalize, or apply what they’ve learned to new situations. The problem with current tests is that some of the examples are actually part of the original training data, which makes it hard to know if the model is really good at detecting out-of-distribution information or just lucky. |
