SIO: Synthetic In-Distribution Data Benefits Out-of-Distribution Detection
This work addresses the problem of OOD detection for machine learning practitioners by providing a plug-and-play method that enhances existing algorithms, though it is incremental as it builds on prior techniques.
The paper tackles the challenge of building reliable Out-of-Distribution (OOD) detectors without requiring external OOD data by using generative models to create synthetic in-distribution images and a novel training objective, resulting in consistent improvements across benchmarks, such as increasing average AUROC from 86.25% to 89.04% on CIFAR-10 vs. CIFAR-100 and achieving a new SOTA of 92.94%.
Building up reliable Out-of-Distribution (OOD) detectors is challenging, often requiring the use of OOD data during training. In this work, we develop a data-driven approach which is distinct and complementary to existing works: Instead of using external OOD data, we fully exploit the internal in-distribution (ID) training set by utilizing generative models to produce additional synthetic ID images. The classifier is then trained using a novel objective that computes weighted loss on real and synthetic ID samples together. Our training framework, which is termed SIO, serves as a "plug-and-play" technique that is designed to be compatible with existing and future OOD detection algorithms, including the ones that leverage available OOD training data. Our experiments on CIFAR-10, CIFAR-100, and ImageNet variants demonstrate that SIO consistently improves the performance of nearly all state-of-the-art (SOTA) OOD detection algorithms. For instance, on the challenging CIFAR-10 v.s. CIFAR-100 detection problem, SIO improves the average OOD detection AUROC of 18 existing methods from 86.25\% to 89.04\% and achieves a new SOTA of 92.94\% according to the OpenOOD benchmark. Code is available at https://github.com/zjysteven/SIO.