STFAR: Improving Object Detection Robustness at Test-Time by Self-Training with Feature Alignment Regularization
This addresses the challenge of domain shift in object detection for real-world applications where target data is only available during inference, representing an incremental advance in test-time adaptation.
The paper tackles the problem of adapting object detection models to unpredictable target domains at test-time, achieving state-of-the-art performance on benchmarks created from three standard datasets.
Domain adaptation helps generalizing object detection models to target domain data with distribution shift. It is often achieved by adapting with access to the whole target domain data. In a more realistic scenario, target distribution is often unpredictable until inference stage. This motivates us to explore adapting an object detection model at test-time, a.k.a. test-time adaptation (TTA). In this work, we approach test-time adaptive object detection (TTAOD) from two perspective. First, we adopt a self-training paradigm to generate pseudo labeled objects with an exponential moving average model. The pseudo labels are further used to supervise adapting source domain model. As self-training is prone to incorrect pseudo labels, we further incorporate aligning feature distributions at two output levels as regularizations to self-training. To validate the performance on TTAOD, we create benchmarks based on three standard object detection datasets and adapt generic TTA methods to object detection task. Extensive evaluations suggest our proposed method sets the state-of-the-art on test-time adaptive object detection task.