Seeing Roads Through Words: A Language-Guided Framework for RGB-T Driving Scene Segmentation
Researchers have developed a new language-guided framework called CLARITY for RGB-Thermal fusion in autonomous driving scene segmentation. This method dynamically adapts its fusion strategy based on detected scene conditions, unlike previous static approaches. CLARITY leverages vision-language model priors to modulate the contribution of each modality and uses object embeddings for segmentation. It also includes mechanisms to preserve valid dark-object semantics and enforce structural consistency across scales, leading to state-of-the-art performance on the MFNet dataset. AI