LM-SCIP uses an LLM with a Channel-Adaptive Semantic Module and H-MoE to perform channel-aware vision-radar fusion, achieving 40% lower localization RMSE than vision-only on nuScenes and strong results on VIRAT.
CRAFT: Camera–Radar 3D Object Detection with Spatio–Contextual Fusion Transformer,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
LLM-Empowered Multimodal Fusion Framework for Autonomous Driving: Semantic Enhancement and Channel-Adaptive Design
LM-SCIP uses an LLM with a Channel-Adaptive Semantic Module and H-MoE to perform channel-aware vision-radar fusion, achieving 40% lower localization RMSE than vision-only on nuScenes and strong results on VIRAT.