NTIRE 2025 跨域少样本目标检测挑战赛：方法与结果
计算机视觉与模式识别
2025-04-16 v1 人工智能
摘要
跨域少样本目标检测（CD-FSOD）在跨域应用时对现有的目标检测与少样本检测模型构成了重大挑战。结合 NTIRE 2025，我们组织了首届 CD-FSOD 挑战赛，旨在推动当前目标检测器在仅有有限标注数据的全新目标域上的性能。本次挑战赛吸引了 152 名注册参与者，收到 42 支队伍的提交，最终有 13 支队伍完成了有效最终提交。参赛者从多元视角处理该任务，提出了在开源与闭源设置下均取得新 SOTA 结果的新模型。在本报告中，我们概述了首届 NTIRE 2025 CD-FSOD 挑战赛，重点介绍了所提出的解决方案并总结了参赛者提交的结果。
引用
@article{arxiv.2504.10685,
title = {NTIRE 2025 Challenge on Cross-Domain Few-Shot Object Detection: Methods and Results},
author = {Yuqian Fu and Xingyu Qiu and Bin Ren and Yanwei Fu and Radu Timofte and Nicu Sebe and Ming-Hsuan Yang and Luc Van Gool and Kaijin Zhang and Qingpeng Nong and Xiugang Dong and Hong Gao and Xiangsheng Zhou and Jiancheng Pan and Yanxing Liu and Xiao He and Jiahao Li and Yuze Sun and Xiaomeng Huang and Zhenyu Zhang and Ran Ma and Yuhan Liu and Zijian Zhuang and Shuai Yi and Yixiong Zou and Lingyi Hong and Mingxi Chen and Runze Li and Xingdong Sheng and Wenqiang Zhang and Weisen Chen and Yongxin Yan and Xinguo Chen and Yuanjie Shao and Zhengrong Zuo and Nong Sang and Hao Wu and Haoran Sun and Shuming Hu and Yan Zhang and Zhiguang Shi and Yu Zhang and Chao Chen and Tao Wang and Da Feng and Linhai Zhuo and Ziming Lin and Yali Huang and Jie Me and Yiming Yang and Mi Guo and Mingyuan Jiu and Mingliang Xu and Maomao Xiong and Qunshu Zhang and Xinyu Cao and Yuqing Yang and Dianmo Sheng and Xuanpu Zhao and Zhiyu Li and Xuyang Ding and Wenqian Li},
journal= {arXiv preprint arXiv:2504.10685},
year = {2025}
}
备注
accepted by CVPRW 25 @ NTIRE