Abdulkarim Atrash · Omar Moured · Yufan Chen · Jiaming Zhang · Seyda Ertekin . Omur Ugur
TY-RIST enhances real-time infrared small target detection (IRSTD) with:
- Stride-aware backbone
- High-resolution detection head
- Cascaded coordinate attention
- Branch pruning
We also employ Normalized Gaussian Wasserstein Distance (NWD) for stable regression.
TY-RIST achieves state-of-the-art results on 4 benchmarks: +7.9% mAP@50 IoU, +3% Precision, +10.2% Recall, and 123 FPS on a single GPU.
To request access to the code in this repository, please complete the form provided below.
@misc{atrash2025tyristtacticalyolotricks,
title={TY-RIST: Tactical YOLO Tricks for Real-time Infrared Small Target Detection},
author={Abdulkarim Atrash and Omar Moured and Yufan Chen and Jiaming Zhang and Seyda Ertekin and Omur Ugur},
year={2025},
eprint={2509.22909},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2509.22909},
}
