Run-Sheng Li

Multimodal UAV Object Detection Using a YOLOv10-Transformer Hybrid Architecture for Balancing Precision and Speed

A global–local Transformer head, alternating depthwise convolutions with sparse self-attention, restores the small-target recall normally lost in attention-heavy designs, while CycleGAN-based cross-modal synthesis and an improved NMS-Free pipeline complete the training and inference path.