CD-FSOD: Cross-Domain Few-Shot Detection

• CD-FSOD is a paradigm that adapts detectors from a richly annotated source domain to a target domain with only a few labeled instances per class while handling significant domain shifts.
• It addresses challenges like overfitting, weak foreground-background separation, and feature confusion through specialized augmentation, fine-tuning, and pseudo-label strategies.
• Recent methods integrate foundation models, multimodal information, and adaptive prototype management to achieve higher mAP across diverse target domains.

Cross-Domain Few-Shot Object Detection (CD-FSOD) studies object detection when a detector must be adapted from a source or base domain to a target domain that differs in visual distribution and offers only a few labeled target instances per class. A common formulation writes a source dataset $\mathcal{D}_S=\{I,y\},\, y\in\mathcal{C}_S$ and a target dataset $\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$, with $\mathcal{C}_S\cap\mathcal{C}_T=\emptyset$ and $\mathcal{P}_S\neq\mathcal{P}_T$ (Fu et al., 2024). In the benchmark setting that established the area, a detector is trained on a base dataset such as MS COCO and then adapted to target domains with $K\in\{1,5,10\}$ annotated instances per class (Xiong, 2022). The problem sits at the intersection of few-shot transfer, domain adaptation, and open-set or open-vocabulary detection, but the conjunction of label scarcity and domain shift produces its own characteristic bottlenecks: overfitting, weak foreground-background separation, feature confusion, and unstable fine-tuning (Xiong, 2022, Meng et al., 2 May 2025, Yu et al., 30 Mar 2026).

1. Formal setting and neighboring problem formulations

In its canonical form, CD-FSOD assumes abundant source-domain supervision and extremely limited target-domain supervision. The detector is first trained on the source domain and then adapted to a target dataset in which each class has only $K$ labeled instances, typically for $K=1,5,10$ (Xiong, 2022). Later challenge formulations make the support/query protocol explicit: the target support set contains $K$ labeled instances per class, and the remaining unlabeled examples form the query set used for evaluation (Fu et al., 14 Apr 2025). An important protocol detail from the NTIRE challenge reports is that “$K$-shot” means $K$ labeled instances, not $\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$0 fully annotated images; support images may therefore contain additional unlabeled target instances, a fact that later became methodologically consequential (Fu et al., 14 Apr 2025).

CD-FSOD differs from standard few-shot object detection because the scarcity of support data is compounded by a genuine distribution shift. It also differs from classical domain-adaptive detection because the target supervision is few-shot rather than abundant or predominantly unlabeled. A stricter adjacent formulation appears in “Few-Shot Object Detection in Unseen Domains,” which treats the setting as zero-shot domain adaptation or domain generalization for FSOD: source-domain few-shot support for novel classes is available, but target-domain images or labels of those novel classes are unavailable during training (Guirguis et al., 2022). This stricter setting is not identical to the now-standard CD-FSOD benchmark, but it clarifies a boundary case in which the domain gap must be handled without any target-domain novel-class supervision.

The field also contains multimodal variants. “Cross-domain Multi-modal Few-shot Object Detection via Rich Text” extends the dataset from $\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$1 to $\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$2, where $\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$3 denotes text associated with object instances or categories, and frames the target problem as CDMM-FSOD (Shangguan et al., 2024). This suggests that CD-FSOD is increasingly understood not only as an image-only transfer problem but also as a setting in which side information, especially language, can act as an auxiliary adaptation signal.

2. Benchmarks, datasets, and evaluation regimes

The first benchmark paper organized CD-FSOD around MS COCO as source data and three target domains: ArTaxOr, UODD, and DIOR. It reports ArTaxOr as a biology dataset with 7 classes, 13,991 train images, and 1,383 test images; UODD as an underwater dataset with 3 classes, 3,194 train images, and 506 test images; and DIOR as an aerial dataset with 20 classes, 18,463 train images, and 5,000 test images (Xiong, 2022). A later benchmark expansion used six target domains—ArTaxOr, Clipart1k, DIOR, DeepFish, NEU-DET, and UODD—and attached to each domain qualitative style labels and quantitative or semi-quantitative indicators of inter-class variance and indefinable boundaries (Fu et al., 2024).

Setting	Domains / datasets	Notes
Initial CD-FSOD benchmark	ArTaxOr, UODD, DIOR	COCO source; 1/5/10-shot; mAP (Xiong, 2022)
Expanded six-domain benchmark	ArTaxOr, Clipart1k, DIOR, DeepFish, NEU-DET, UODD	style, ICV, and IB analysis (Fu et al., 2024)
NTIRE 2025 challenge	Development on six domains above; final unseen DeepFruits, CARPK, CarDD	open-source and closed-source tracks (Fu et al., 14 Apr 2025)
NTIRE 2026 challenge	second public challenge under open-source and closed-source tracks	128 registered participants and 696 submissions (Qiu et al., 13 Apr 2026)

Evaluation metrics are not entirely uniform across the literature. The benchmark paper reports mAP (Xiong, 2022). CD-ViTO’s benchmark tables also report mAP on the six target domains (Fu et al., 2024). CDFormer explicitly reports [email protected]:0.95 (Meng et al., 2 May 2025), while the NTIRE reports use mAP as the ranking metric and weight 1-shot performance more strongly in the final challenge score (Fu et al., 14 Apr 2025). This implies that reported numbers are most reliable when interpreted within the protocol of each paper rather than as unqualified cross-paper absolutes.

The challenge reports also enlarged the benchmarking function of CD-FSOD. The 2025 NTIRE challenge introduced open-source and closed-source tracks, used the six benchmark domains for development, and evaluated on three unseen final datasets—DeepFruits, CARPK, and CarDD—under 1-shot, 5-shot, and 10-shot conditions (Fu et al., 14 Apr 2025). The 2026 NTIRE challenge continued the two-track structure and reported 128 registered participants, 696 submissions, 31 active teams, and 19 valid final results, underscoring the topic’s transition from a small benchmark niche to a public evaluation ecosystem (Qiu et al., 13 Apr 2026).

3. Diagnosed failure modes and analytic frameworks

The earliest benchmark results established that same-domain FSOD intuitions transfer poorly to CD-FSOD. On the initial benchmark, tested meta-learning methods such as A-RPN, Meta-RCNN, and H-GCN underperformed or only marginally matched a naive Faster R-CNN fine-tuning baseline, while freezing parameters consistently hurt performance (Xiong, 2022). The benchmark paper explicitly attributed this to overfitting and to the inability of source-domain episodic training to mimic distant target domains (Xiong, 2022). This remains a foundational claim in the area: the central difficulty is not only data scarcity, but the interaction between scarcity and distributional mismatch.

CD-ViTO reframed the domain gap through three axes: style, inter-class variance (ICV), and indefinable boundaries (IB) (Fu et al., 2024). ICV is computed from CLIP text features via

$\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$4

and was reported as small for ArTaxOr and UODD, medium for DIOR, and large for Clipart1k and NEU-DET (Fu et al., 2024). IB was estimated through a user study and summarized as

$\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$5

with ArTaxOr, Clipart1k, and DIOR labeled slight, DeepFish moderate, and NEU-DET and UODD significant (Fu et al., 2024). In that analysis, style alone did not explain the full degradation; small ICV and especially large IB were more predictive of poor transfer (Fu et al., 2024).

Later work moved from dataset-level diagnostics to representation-level failure modes. CDFormer argues that the decisive problem is feature confusion, divided into object-background confusion and object-object confusion, and introduces separate modules for each (Meng et al., 2 May 2025). A still more recent analysis identifies a “target-domain Astigmatism” phenomenon in transformer detectors: attention becomes dispersed and unfocused in the target domain, producing imprecise localization and redundant predictions, while regular fine-tuning only partially remedies this effect (Jiang et al., 19 Mar 2026). A plausible implication is that CD-FSOD has progressed from being viewed mainly as a sample-efficiency problem to being studied as a problem of representation geometry, attention allocation, and background modeling under shift.

Fine-tuning behavior itself has become an analytical object. “A Closer Look at Cross-Domain Few-Shot Object Detection: Fine-Tuning Matters and Parallel Decoder Helps” explicitly identifies unstable optimization and weak generalization during adaptation as primary bottlenecks and evaluates robustness on an OOD-contaminated mixed-domain test set built from CD-FSOD (Yu et al., 30 Mar 2026). This line of work suggests that CD-FSOD performance is sensitive not only to what representations are used, but also to how adaptation is staged and regularized.

4. Methodological development

The earliest methods in and around CD-FSOD were shaped by augmentation, selective adaptation, and regularization. “Few-Shot Object Detection in Unseen Domains” proposed a two-fold strategy: learn domain shift on base classes through meta-training and augment few-shot novel examples with color jittering, Gaussian blur, Gaussian noise, background replacement, and feature-level sampling, plus a contrastive foreground-class embedding loss to reduce background bias (Guirguis et al., 2022). AcroFOD formalized an adaptive optimization view of augmentation: generate cross-domain candidates, score them by distance to target features, and keep only the closest fraction under a shrinkage ratio $\mathcal{D}_T=\{I,y\},\, y\in\mathcal{C}_T$6; it combined this with image-level and box-level domain-aware augmentation (Gao et al., 2022). The original CD-FSOD benchmark paper contributed an EMA teacher-student baseline with pseudo-label distillation, emphasizing overfitting control rather than complex source-task design (Xiong, 2022).

A second phase of the literature reoriented the problem around open-set and open-vocabulary detectors. CD-ViTO showed that DE-ViT, despite strong same-domain FSOD results, generalized poorly to CD-FSOD, and addressed this with learnable instance features, instance reweighting, and a domain prompter (Fu et al., 2024). The multimodal extension “Cross-domain Multi-modal Few-shot Object Detection via Rich Text” argued that rich textual descriptions could stabilize support-query transfer under severe domain shift; it fused support visual and text features and used a bidirectional rich text rectification loss (Shangguan et al., 2024). This line of work made two claims that continued to recur: first, language can supply domain-invariant semantics; second, semantics alone are not sufficient when localization depends on target-domain appearance.

Recent work has been dominated by foundation-model adaptation. “Enhance Then Search” treats CD-FSOD as a search problem over augmentation policies and fine-tuning configurations on top of GroundingDINO, using a mixed augmentation pipeline with CachedMosaic, YOLOXHSVRandomAug, RandomFlip, CachedMixUp, RandomResize, and RandomCrop, plus validation-guided grid search (Pan et al., 6 Apr 2025). Domain-RAG shifts attention to training-free generative augmentation: it preserves the foreground object, retrieves domain-aware backgrounds using CLIP and shallow ResNet statistics, synthesizes new backgrounds with FLUX-based generation, and recomposes support images, yielding a “fix the foreground, adapt the background” strategy (Li et al., 6 Jun 2025). GiPL combines iterative pseudo-label self-training on the support set with generative data augmentation using large vision-LLMs to enrich training samples and suppress overfitting (Liu et al., 28 May 2026). Challenge-winning systems further amplified the role of self-training and annotation completion, often because support images in the challenge protocol were only partially annotated (Fu et al., 14 Apr 2025).

A parallel thread has concentrated on representation design rather than data generation. CDFormer uses a DETR-style single-stage architecture with an Object-Background Distinguishing module, driven by a learnable background token, and an Object-Object Distinguishing module with an InfoNCE objective over learned class embeddings (Meng et al., 2 May 2025). FSOD-VFM discards fine-tuning entirely: it combines a Universal Proposal Network, SAM2 masks, DINOv2 features, and graph-based confidence diffusion to refine fragmented proposals, and reports strong training-free CD-FSOD performance (Feng et al., 3 Feb 2026). LMP returns to multimodality but with a different decomposition from text-enrichment work: it builds class-level visual prototypes from support RoIs, generates hard-negative prototypes from jittered query boxes, and couples a visual-guided branch to a parallel text-guided branch on top of GroundingDINO (Wang et al., 21 Feb 2026). The fine-tuning-centric line represented by the hybrid ensemble decoder argues that stronger adaptation can be obtained without elaborate synthetic pipelines, by combining a shared hierarchical decoder layer with multiple parallel branches and a progressive freeze-then-unfreeze schedule (Yu et al., 30 Mar 2026). The newest attention-centric proposal frames target-domain error as a center-periphery failure and introduces Positive Pattern Refinement, Negative Context Modulation, and Textual Semantic Alignment to convert “astigmatic” attention into more focused patterns (Jiang et al., 19 Mar 2026).

5. Empirical progression and challenge outcomes

The initial benchmark established a low baseline ceiling. Averaged across ArTaxOr, UODD, and DIOR, A-RPN reached 10.2 mAP, Meta-RCNN 10.6, H-GCN 10.5, FRCN-ft 11.2, and the benchmark’s EMA teacher-student method 13.8, yielding a reported average gain of 2.0 mAP over the second-best approach (Xiong, 2022). This result was important less for its absolute number than for its conclusion: CD-FSOD did not behave like standard FSOD, and much of the design space inherited from VOC/COCO few-shot detection was inadequate (Xiong, 2022).

On the six-domain benchmark, CD-ViTO raised average mAP from DE-ViT’s 1.0/5.1/5.9 to 13.9/26.1/29.6 for 1/5/10-shot, respectively (Fu et al., 2024). CDFormer then reported 26.8/37.1/40.0 average mAP and improvements of 12.9, 11.0, and 10.4 mAP over the previous best under 1/5/10-shot settings (Meng et al., 2 May 2025). Domain-RAG, built on GroundingDINO, reported 33.6/42.7/45.4 average mAP and gains of +7.3, +2.3, and +1.5 over the corresponding GroundingDINO baseline (Li et al., 6 Jun 2025). LMP reported 34.3/44.0/46.6 average mAP on the same six-domain benchmark, improving over fine-tuned GroundingDINO from 26.3/40.4/43.9 (Wang et al., 21 Feb 2026). The fine-tuning-centric hybrid ensemble decoder further reported 34.9/45.0/47.9 on CD-FSOD, together with gains on ODinW-13 and RF100-VL, indicating that careful adaptation of pretrained open-vocabulary detectors remains highly competitive (Yu et al., 30 Mar 2026).

Training-free performance has also improved materially. FSOD-VFM reported 25.3/31.3/31.6 nAP on CD-FSOD for 1/5/10-shot and emphasized that its 10-shot result of 31.6 AP substantially exceeded previous training-free methods reported at 21.4 AP (Feng et al., 3 Feb 2026). This does not overtake the best fine-tuned open-source systems, but it demonstrates that part of the CD-FSOD difficulty can be shifted from optimization to proposal quality, foreground masking, and prototype matching (Feng et al., 3 Feb 2026).

The NTIRE challenges clarified larger ecosystem trends. The 2025 report states that top open-source methods were dominated by foundation-model transfer, self-training or pseudo-label completion, augmentation, and, in some cases, ensembling of GLIP and GroundingDINO variants (Fu et al., 14 Apr 2025). The winning open-source system, MoveFree, combined Grounding DINO, self-training, and Mixture-of-Experts inside the decoder; AI4EarthLab’s ETS ranked second; IDCFS combined GLIP, iterative pseudo-labeling, LoRA-adapted GroundingDINO, and ensembling (Fu et al., 14 Apr 2025). In the closed-source track, prototype and memory mechanisms remained central: X-Few’s Instance Feature Caching topped the track with a score of 125.90 against a CD-ViTO baseline of 91.00 (Fu et al., 14 Apr 2025). This suggests a bifurcation of the field: in open-source settings, large pretrained grounding detectors and pseudo-label refinement dominate; in closed-source settings, prototype calibration, caching, and feature refinement remain the main levers.

6. Open questions, controversies, and future directions

One recurring controversy concerns what exactly is being measured in open-source CD-FSOD. Challenge reports explicitly note that once large foundation models are permitted, strict non-overlap between source and target classes becomes difficult to guarantee (Fu et al., 14 Apr 2025). This does not invalidate the task, but it changes its interpretation: open-source CD-FSOD often becomes a problem of few-shot specialization under domain shift rather than a strict disjoint-class transfer experiment. A related ambiguity appears in support supervision itself: because challenge support images may contain unlabeled target objects, pseudo-label completion can improve performance partly by repairing annotation incompleteness rather than only by learning better detectors (Fu et al., 14 Apr 2025). This suggests that evaluation protocols and annotation assumptions are not neutral implementation details; they actively shape which methods look best.

A second tension concerns fine-tuning versus training-free adaptation. The strongest averages on current six-domain benchmarks come from careful fine-tuning of large pretrained detectors (Yu et al., 30 Mar 2026, Wang et al., 21 Feb 2026), yet FSOD-VFM shows that training-free pipelines can already be competitive, especially when proposal generation, masking, and graph-based reweighting are strong (Feng et al., 3 Feb 2026). This suggests that the field is not converging on a single paradigm. Instead, it is separating into optimization-heavy and optimization-light branches, each with different assumptions about compute budget, robustness, and deployment constraints.

A third unresolved issue is how best to use multimodality. Rich text improves CDMM-FSOD substantially in severe low-shot regimes, but richer text is not automatically better: on NEU-DET, LLM-generated descriptions underperformed manually designed text (Shangguan et al., 2024). LMP, by contrast, argues that the missing ingredient in text-dominant open-vocabulary detectors is domain-specific visual information, especially hard negatives (Wang et al., 21 Feb 2026). This suggests a broader lesson: language is valuable as a semantic prior, but CD-FSOD remains a localization problem, and localization under domain shift often depends on target-domain appearance cues that text alone does not encode.

The persistent difficulty of high-IB domains such as NEU-DET and UODD also remains unresolved (Fu et al., 2024). Across multiple papers, these datasets continue to expose failures in background suppression, precise boundary modeling, and fine-grained support quality assessment (Fu et al., 2024, Meng et al., 2 May 2025). Recent attention-based analyses and center-periphery refinement mechanisms imply that future progress may come from finer control of where detectors attend in target-domain images rather than from larger support sets alone (Jiang et al., 19 Mar 2026).

The present trajectory of the field points toward several concrete research directions. One is better OOD-aware evaluation and adaptation, motivated by mixed-domain tests such as CD-Mixed (Yu et al., 30 Mar 2026). Another is more principled support-set completion and pseudo-label refinement for partially annotated few-shot images (Fu et al., 14 Apr 2025). A third is adaptive prototype management—selection, pruning, negative mining, and multimodal fusion—which appears across LMP, rich-text CDMM-FSOD, and closed-source challenge methods (Wang et al., 21 Feb 2026, Shangguan et al., 2024, Fu et al., 14 Apr 2025). A plausible synthesis is that future CD-FSOD systems will combine robust pretrained grounding detectors, explicit target-domain prototype construction, selective pseudo-label expansion, and finer-grained attention control, while evaluation protocols continue to distinguish more clearly between closed-source scientific comparison and open-source practical transfer.