Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

NTIRE 2025 Challenge on Cross-Domain Few-Shot Object Detection: Methods and Results

Yuqian Fu, Xingyu Qiu, Bin Ren, Yanwei Fu, Radu Timofte, Nicu Sebe, Ming-Hsuan Yang, Luc Van Gool, Kaijin Zhang, Qingpeng Nong, Xiugang Dong, Hong Gao, Xiangsheng Zhou, Jiancheng Pan, Yanxing Liu, Xiao He, Jiahao Li, Yuze Sun, Xiaomeng Huang, Zhenyu Zhang, Ran Ma, Yuhan Liu, Zijian Zhuang, Shuai Yi, Yixiong Zou, Lingyi Hong, Mingxi Chen, Runze Li, Xingdong Sheng, Wenqiang Zhang, Weisen Chen, Yongxin Yan, Xinguo Chen, Yuanjie Shao, Zhengrong Zuo, Nong Sang, Hao Wu, Haoran Sun, Shuming Hu, Yan Zhang, Zhiguang Shi, Yu Zhang, Chao Chen, Tao Wang, Da Feng, Linhai Zhuo, Ziming Lin, Yali Huang, Jie Me, Yiming Yang, Mi Guo, Mingyuan Jiu, Mingliang Xu, Maomao Xiong, Qunshu Zhang, Xinyu Cao, Yuqing Yang, Dianmo Sheng, Xuanpu Zhao, Zhiyu Li, Xuyang Ding, Wenqian Li

TL;DR

This paper introduces the NTIRE 2025 CD-FSOD Challenge, a benchmark and evaluation framework for Cross-Domain Few-Shot Object Detection with open-source and closed-source tracks. It builds on the CD-ViTO baseline and showcases a diverse set of methods that combine vision-language grounding, open-set detectors, domain augmentation, pseudo-labeling, and memory-based adaptation to address domain gaps and data scarcity. The results demonstrate tangible gains by leveraging foundation models, self-training, and domain-adaptive components, validating the practicality of cross-domain few-shot detection in varied real-world domains. The work establishes a rigorous, multi-dataset, multi-shot evaluation protocol and highlights promising directions for robust CD-FSOD development and fair cross-team comparisons.

Abstract

Cross-Domain Few-Shot Object Detection (CD-FSOD) poses significant challenges to existing object detection and few-shot detection models when applied across domains. In conjunction with NTIRE 2025, we organized the 1st CD-FSOD Challenge, aiming to advance the performance of current object detectors on entirely novel target domains with only limited labeled data. The challenge attracted 152 registered participants, received submissions from 42 teams, and concluded with 13 teams making valid final submissions. Participants approached the task from diverse perspectives, proposing novel models that achieved new state-of-the-art (SOTA) results under both open-source and closed-source settings. In this report, we present an overview of the 1st NTIRE 2025 CD-FSOD Challenge, highlighting the proposed solutions and summarizing the results submitted by the participants.

NTIRE 2025 Challenge on Cross-Domain Few-Shot Object Detection: Methods and Results

TL;DR

This paper introduces the NTIRE 2025 CD-FSOD Challenge, a benchmark and evaluation framework for Cross-Domain Few-Shot Object Detection with open-source and closed-source tracks. It builds on the CD-ViTO baseline and showcases a diverse set of methods that combine vision-language grounding, open-set detectors, domain augmentation, pseudo-labeling, and memory-based adaptation to address domain gaps and data scarcity. The results demonstrate tangible gains by leveraging foundation models, self-training, and domain-adaptive components, validating the practicality of cross-domain few-shot detection in varied real-world domains. The work establishes a rigorous, multi-dataset, multi-shot evaluation protocol and highlights promising directions for robust CD-FSOD development and fair cross-team comparisons.

Abstract

Cross-Domain Few-Shot Object Detection (CD-FSOD) poses significant challenges to existing object detection and few-shot detection models when applied across domains. In conjunction with NTIRE 2025, we organized the 1st CD-FSOD Challenge, aiming to advance the performance of current object detectors on entirely novel target domains with only limited labeled data. The challenge attracted 152 registered participants, received submissions from 42 teams, and concluded with 13 teams making valid final submissions. Participants approached the task from diverse perspectives, proposing novel models that achieved new state-of-the-art (SOTA) results under both open-source and closed-source settings. In this report, we present an overview of the 1st NTIRE 2025 CD-FSOD Challenge, highlighting the proposed solutions and summarizing the results submitted by the participants.

Paper Structure

This paper contains 50 sections, 17 equations, 19 figures, 2 tables.

Table of Contents

  1. Introduction
  2. NTIRE 2025 CD-FSOD Challenge
  3. Challenge Overview
  4. Task Formulations
  5. Challenge Phases and Datasets
  6. CD-ViTO Baseline Model
  7. Evaluation Protocol
  8. Challenge Results
  9. Main Open-Source Track Methods
  10. MoveFree
  11. Proposed Method
  12. Training Details
  13. AI4EarthLab
  14. Proposed Method
  15. Training Details
  16. ...and 35 more sections

Figures (19)

  • Figure 1: Illustration of the challenge settings, including the closed-source and open-source CD-FSOD tracks. The three newly introduced target datasets used in the final testing phase are also shown.
  • Figure 2: Overall framework of CD-ViTO baseline method.
  • Figure 3: Team MoveFree: an illustration of the substitution of MoE into Grounding DINO's decoder layers.
  • Figure 4: Team AI4EarthLab: overall framework of augmentation-search strategy Enhance Then Search (ETS) with foundation model for CD-FSOD.
  • Figure 5: Team IDCFS: overview of the proposed Pseudo-Label Driven Vision-Language Grounding for CD-FSOD.
  • ...and 14 more figures