An Adverse Weather-Immune Scheme with Unfolded Regularization and Foundation Model Knowledge Distillation for Street Scene Understanding

Various adverse weather conditions pose a significant challenge to autonomous driving (AD) perception. A common strategy is to minimize the disparity between images captured in clear and adverse weather conditions. However, this technique typically relies on utilizing clear image as a reference, which is challenging to obtain in practice. Furthermore, this method typically targets a single adverse condition and perform poorly when confronting the mixup of multiple adverse weather conditions. To address these issues, we introduce a reference-free and \underline{Adv}erse weather-\underline{Immu}ne scheme (called AdvImmu) achieved by leveraging the invariance of weather conditions over short periods (seconds). Specifically, AdvImmu includes three components: Locally Sequential Mechanism (LSM), Globally Shuffled Mechanism (GSM), and Unfolded Regularizers (URs). LSM leverages temporal correlations between adjacent frames to enhance model performance. GSM is proposed to shuffle LSM segments to prevent the overfitting to temporal patterns of only using LSM. URs are the deep unfolding implementation of two proposed regularizers to penalize the model complexity to enhance across-weather generalization. In addition, to overcome the over-reliance on consecutive frame-wise annotations in the training of AdvImmu (typically unavailable in AD scenarios), we incorporate the Segment Anything Model (SAM) to annotate frames, and additionally propose a cluster algorithm (denoted as SBICAC) to surmount SAM's category-agnostic issue to generate pseudo-labels. Extensive experiments demonstrate that the proposed AdvImmu outperforms existing state-of-the-art methods by 88.56\% in mean Intersection over Union (mIoU).
Comment: 10 Pages