Raising the Bar on the Evaluation of Out-of-Distribution Detection
This work addresses the evaluation gap in OoD detection for image classification, which is incremental as it refines existing benchmarks rather than introducing a new detection method.
The paper tackles the problem of evaluating out-of-distribution (OoD) detection methods by defining two categories of OoD data based on perceptual and semantic similarity, and proposes a GAN-based framework to generate such samples. It shows that state-of-the-art methods perform significantly worse on this new benchmark, with models performing well on it also doing well on conventional benchmarks, suggesting a unified evaluation approach.
In image classification, a lot of development has happened in detecting out-of-distribution (OoD) data. However, most OoD detection methods are evaluated on a standard set of datasets, arbitrarily different from training data. There is no clear definition of what forms a ``good" OoD dataset. Furthermore, the state-of-the-art OoD detection methods already achieve near perfect results on these standard benchmarks. In this paper, we define 2 categories of OoD data using the subtly different concepts of perceptual/visual and semantic similarity to in-distribution (iD) data. We define Near OoD samples as perceptually similar but semantically different from iD samples, and Shifted samples as points which are visually different but semantically akin to iD data. We then propose a GAN based framework for generating OoD samples from each of these 2 categories, given an iD dataset. Through extensive experiments on MNIST, CIFAR-10/100 and ImageNet, we show that a) state-of-the-art OoD detection methods which perform exceedingly well on conventional benchmarks are significantly less robust to our proposed benchmark. Moreover, b) models performing well on our setup also perform well on conventional real-world OoD detection benchmarks and vice versa, thereby indicating that one might not even need a separate OoD set, to reliably evaluate performance in OoD detection.