Activate and Reject: Towards Safe Domain Generalization under Category Shift

Albeit the notable performance on in-domain test points, it is non-trivial for deep neural networks to attain satisfactory accuracy when deploying in the open world, where novel domains and object classes often occur. In this paper, we study a practical problem of Domain Generalization under Category Shift (DGCS), which aims to simultaneously detect unknown-class samples and classify known-class samples in the target domains. Compared to prior DG works, we face two new challenges: 1) how to learn the concept of "unknown " during training with only source known- class samples, and 2) how to adapt the source-trained model to unseen environments for safe model deployment. To this end, we propose a novel Activate and Reject (ART) framework to reshape the model’s decision boundary to accommodate unknown classes and conduct post hoc modification to further discriminate known and unknown classes using unlabeled test data. Specifically, during training, we promote the response to the unknown by optimizing the unknown probability and then smoothing the overall output to mitigate the overconfidence issue. At test time, we introduce a step-wise online adaptation method that predicts the label by virtue of the cross-domain nearest neighbor and class prototype information without updating the network’s parameters or using threshold-based mechanisms. Experiments reveal that ART consistently improves the generalization capability of deep networks on different vision tasks. For image classification, ART improves the H-score by 6.1% on average compared to the previous best method. For object detection and semantic segmentation, we establish new benchmarks and achieve competitive performance.

1. Introduction

Deep neural networks have achieved unprecedented success in a myriad of vision tasks over the past decade. Despite the promise, a well-trained model deployed in the open and ever-changing world often struggles to deal with the domain shifts—the training and testing data do not follow the independent and identically distributed (i.i.d) assumption, and therefore deteriorates its safety and reliability in many safety-critical applications, such as autonomous driving and computer-aided disease diagnosis. This gives rise to the importance of Domain Generalization (DG) [101], [83], a.k.a. out-of-distribution (OOD) generalization, which aims at generalizing predictive models trained on multiple (or a single) source domains to unseen target distributions.