TY-RIST: Tactical YOLO Tricks for Real-Time Infrared Small Target Detection

While critical for defense and surveillance, infrared small target detection (IRSTD) remains a challenging task due to: (1) target loss from minimal features, (2) false alarms in cluttered environments, (3) missed detections from low saliency, and (4) high computational costs. To address these, we propose TY-RIST, an optimized YOLOv12n architecture featuring: (1) a stride-aware backbone with fine-grained receptive fields, (2) a high-resolution detection head, (3) cascaded coordinate attention blocks, and (4) a branch pruning strategy that reduces computational cost up to ~25.5% while marginally enhancing performance and enabling real-time inference. Additionally, we incorporate the Normalized Gaussian Wasserstein Distance (NWD) to improve regression stability. Extensive experiments on four benchmarks and across 20 different models demonstrate state-of-the-art performance, boosting mAP@50 by + 7.9%, Precision by +3%, and Recall by + 10.2%, while running up to ~123 FPS on a single GPU. Cross-dataset validation on a fifth dataset further confirms strong generalization capability. Further results and details are published at www.github.com/moured/TY-RIST.