Cross Modality YOLO: Object Detection Using RGB-IR Images in Low Visibility Conditions

Object detection in low visibility conditions is crucial for applications such as autonomous driving, surveillance, and military operations. Although deep learning based detectors have advanced markedly, conventional RGB only models remain vulnerable to challenging scenarios such as darkness, fog, and strong backlight. To overcome these limitations, we propose CMYOLO, a cross modality object detector that integrates a Transformer-based cross attention fusion mechanism with dual feature-enhancement modules. Experiments on the LLVIP dataset show that CMYOLO achieves a higher mAP (mean Average Precision) than existing methods. These results demonstrate that CMYOLO offers superior performance and robustness under low visibility conditions.