Generalizable Fourier Augmentation for Unsupervised Video Object Segmentation

The performance of existing unsupervised video object segmentation methods typically suffers from severe performance degradation on test videos when tested in out-of-distribution scenarios. The primary reason is that the test data in realworld may not follow the independent and identically distribution (i.i.d.) assumption, leading to domain shift. In this paper, we propose a Generalizable Fourier Augmentation (G-FA) method during training to improve the generalization ability of the model. To achieve this, the GFA performs Fast Fourier Transform (FFT) over the intermediate spatial domain features in each layer to yield corresponding frequency representations, including amplitude components (encoding scene-aware styles such as texture, color, contrast of the scene) and phase components (encoding rich semantics). We produce a variety of style features via Gaussian sampling to augment the training data, thereby improving the generalization capability of the model. To further improve the crossdomain generalization performance of the model, we design a phase feature update strategy via exponential moving average using phase features from past frames in an online update manner, which could help the model to learn cross-domaininvariant features. Extensive experiments show that the proposed GFA achieves the state-of-the-art performance on popular benchmarks.