Open-vocabulary object detectors enable prompt-driven recognition, yet their zero-shot transfer to unmanned aerial vehicle (UAV) imagery remains fragile under domain shift, where tiny and cluttered targets depend on weak fine-grained cues. We propose AeroPinWorld, a pinwheel-augmented YOLO-World v2 that revisits stride-2 downsampling as a key transfer bottleneck: aggressive resolution reduction can induce aliasing-driven detail loss and sampling-phase sensitivity, which disproportionately harms small-object representations and degrades cross-dataset generalization in aerial scenes. To address this, AeroPinWorld introduces pinwheel-shaped convolution (PConv) as a phase-aware reduction operator. PConv probes complementary offsets via asymmetric padding and directional kernels before feature fusion, strengthening local structure aggregation at downsampling junctions. Importantly, we do not replace all downsampling operations; instead, we selectively substitute PConv at critical pyramid transitions, including the first two backbone reductions (P1/2 and P2/4) and the two bottom-up stride-2 reductions in the head, while keeping later backbone stages unchanged to preserve efficiency. We evaluate under a strict zero-shot cross-dataset protocol by training on COCO2017 for 24 epochs from official pretrained weights and directly testing on two UAV benchmarks, VisDrone2019-DET and UAVDT, without any target-domain fine-tuning, using an offline prompt vocabulary at inference. Experiments demonstrate consistent improvements over the baseline, achieving a +2.3 mAP gain and a +0.9 APS gain on VisDrone, and yielding consistent gains on UAVDT, while maintaining a competitive efficiency profile.

Few-shot learning (FSL) and zero-shot learning (ZSL) are usually studied as separate problems, yet both require prediction when class-specific evidence is absent or scarce. This review analyzes their shared difficulty from an optimization perspective. Instead of grouping studies only by architecture, it tracks four common coordinates: the information available to the learner, the variables estimated from that information, the objectives and constraints, and the numerical solvers. These coordinates support a unified comparison of attribute-based ZSL, episodic meta-learning, metric and prototype estimators, graph and optimal-transport inference, generative any-shot models, and adaptation of vision–language models. The synthesis exposes recurring trade-offs rather than a universally preferable family: flexible updates increase estimator variance; tractable task-time solvers inherit representation bias; query batches can improve inference while changing the protocol; and strong pretrained priors reduce target-data requirements while making the origin of task evidence harder to audit. Canonical objectives are distinguished from simplified review formulations and prospective research targets. The framework also clarifies the progression from explicit semantic mappings to local adaptation around pretrained image–text representations. Three priorities emerge: model selection without extra validation labels, safe use of uncertain pretrained knowledge, and stable parameter-efficient adaptation. Under this view, FSL and ZSL are connected structured-estimation problems rather than an inventory of unrelated algorithms.

Zero-shot learning (ZSL) addresses scarce samples and unseen categories in industrial defect detection. Zero-shot multi-type anomaly segmentation (MTAS) demands both anomaly localization and fine-grained defect classification, yet mainstream methods like MultiADS face an inherent alignment-generalization paradox: they rely on static prompts and one-way alignment, failing to adapt to fine-grained visual variations and limiting accuracy.To tackle this, we propose the Dynamic Anomaly-Aware Segmentation (DAAS) framework for zero-shot industrial defect detection. Its core is a customized training-only Post-VCP module. Based on industrial defect-tailored multi-head cross-attention over local visual features, this module provides strong supervision for text embedding learning, preserves semantic priors via residual connections, and enhances the fine-grained modeling ability of static text embeddings for inference.Extensive transductive experiments on four benchmarks (MVTec-AD, VisA, MPDD, MAD-real) show DAAS achieves substantial gains on MTAS over MultiADS, with competitive results on binary ZSL. These validate our method’s effectiveness, theoretical rationality and robustness in real industrial scenarios.

Open-vocabulary object detection in unmanned aerial vehicle (UAV) imagery remains challenging under zero-shot cross-dataset transfer because tiny and cluttered targets are highly sensitive to early resolution reduction under domain shift. This study aims to improve transferable open-vocabulary UAV detection by revisiting stride-2 downsampling in YOLO-World v2 as a transfer-critical bottleneck. AeroPinWorld is introduced as a pinwheel-augmented YOLO-World v2 that selectively replaces four key stride-2 transitions with pinwheel-shaped convolution (PConv) so as to reduce aliasing, alleviate sampling-phase sensitivity, and preserve fine-grained local structures, while keeping the original detection head unchanged to ensure a fair and efficient comparison. The model is trained on COCO2017 for 24 epochs from official pretrained weights and directly evaluated, without target-domain fine-tuning, on VisDrone2019-DET and UAVDT using fixed offline prompt vocabularies. Compared with YOLO-World v2-S, AeroPinWorld improves zero-shot transfer performance on VisDrone from 0.112 to 0.135 mAP and from 0.054 to 0.063 AP_S, and it also yields consistent gains on UAVDT. Ablation studies show that both early backbone replacements and head bottom–up replacements contribute to the final gains, with their combination achieving the best accuracy–efficiency trade-off. These results indicate that selectively redesigning transfer-critical downsampling operators is an effective and lightweight way to improve zero-shot open-vocabulary UAV detection under aerial domain shift.