NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation
This addresses a critical limitation for real-world applications like autonomous driving where test data is often non-i.i.d., though it is an incremental improvement over existing TTA methods.
The paper tackles the problem of test-time adaptation (TTA) failing under temporally correlated (non-i.i.d.) test data streams, such as in autonomous driving, by proposing a robust TTA scheme that outperforms state-of-the-art methods in non-i.i.d. settings and matches their performance under i.i.d. assumptions.
Test-time adaptation (TTA) is an emerging paradigm that addresses distributional shifts between training and testing phases without additional data acquisition or labeling cost; only unlabeled test data streams are used for continual model adaptation. Previous TTA schemes assume that the test samples are independent and identically distributed (i.i.d.), even though they are often temporally correlated (non-i.i.d.) in application scenarios, e.g., autonomous driving. We discover that most existing TTA methods fail dramatically under such scenarios. Motivated by this, we present a new test-time adaptation scheme that is robust against non-i.i.d. test data streams. Our novelty is mainly two-fold: (a) Instance-Aware Batch Normalization (IABN) that corrects normalization for out-of-distribution samples, and (b) Prediction-balanced Reservoir Sampling (PBRS) that simulates i.i.d. data stream from non-i.i.d. stream in a class-balanced manner. Our evaluation with various datasets, including real-world non-i.i.d. streams, demonstrates that the proposed robust TTA not only outperforms state-of-the-art TTA algorithms in the non-i.i.d. setting, but also achieves comparable performance to those algorithms under the i.i.d. assumption. Code is available at https://github.com/TaesikGong/NOTE.