Heuristic Self-Paced Learning for Domain Adaptive Semantic Segmentation under Adverse Conditions

ArXi:2603.24322v1 Announce Type: new The learning order of semantic classes significantly impacts unsupervised domain adaptation for semantic segmentation, especially under adverse weather conditions. Most existing curricula rely on handcrafted heuristics (e.g., fixed uncertainty metrics) and follow a static schedule, which fails to adapt to a model's evolving, high-dimensional