Real-time object detection is critical for unmanned aerial vehicles (UAVs) performing various tasks. However, efficiently deploying detection models on UAV platforms with limited storage and computational resources remains a significant challenge. To address this issue, we propose ELNet, an efficient and lightweight object detection model based on YOLOv12n. First, based on an analysis of UAV image characteristics, we strategically remove two A2C2f modules from YOLOv12n and adjust the size and number of detection heads. Second, we propose a novel lightweight detection head, EPGHead, to alleviate the computational burden introduced by adding the large-scale detection head. In addition, since YOLOv12n employs standard convolution for downsampling, which is inefficient for extracting UAV image features, we design a novel downsampling module, EDown, to further reduce model size and enable more efficient feature extraction. Finally, to improve detection in UAV imagery with dense, small, and scale-varying objects, we propose DIMB-C3k2, an enhanced module built upon C3k2, which boosts feature extraction under complex conditions. Compared with YOLOv12n, ELNet achieves an 88.5% reduction in parameter count and a 52.3% decrease in FLOPs, while increasing mAP₅₀ by 1.2% on the VisDrone dataset and 0.8% on the HIT-UAV dataset, reaching 94.7% mAP₅₀ on HIT-UAV. Furthermore, the model achieves a frame rate of 682 FPS, highlighting its superior computational efficiency without sacrificing detection accuracy.

RGB-Thermal (RGB-T) object detection in Unmanned Aerial Vehicle (UAV) imagery typically relies on fusing multimodal information to detect multiple objects under all-weather conditions. However, existing fusion methods primarily rely on pairwise attention mechanisms, thereby overlooking group-wise dependencies. Such oversight renders these methods ineffective at capturing collective relationships among objects in complex aerial scenes, including vehicle queues or cases of heavy occlusion. Additionally, the densely connected structures in these fusion methods often result in high computational costs. To address these limitations, we propose a lightweight cross-modal hypergraph fusion for RGBT object detection, termed HyperDet. The core of HyperDet is an Adaptive Hierarchical Multi-Modal Fusion (AHMF) framework, which follows a cascaded enhancement-reasoning paradigm. Specifically, AHMF first employs a Dual Cross-Modal Feature Alignment (DCFA) module to decouple and align dense RGB textures and sparse thermal signatures, producing a calibrated feature space. On top of this aligned representation, a Cross-Modal Adaptive Hypergraph (CAH) module is introduced to perform high-order reasoning. Unlike pairwise approaches, CAH constructs dynamic hyperedges that aggregate spatially dispersed but semantically consistent instances across modalities. This mechanism enables the model to explicitly capture long-range, group-wise dependencies among objects, allowing for robust recognition of complex clusters. Extensive experiments on three RGBT object detection benchmarks, i.e., DroneVehicle, VEDAI, and LLVIP, demonstrate that HyperDet achieves state-of-the-art performance. For example, it achieves 86.4% mAP\(_{50}\) on DroneVehicle with only 3.2 M parameters. The code for HyperDet is available at https://github.com/lihui-ioe/HyperDet.

Object detection in Unmanned Aerial Vehicle (UAV) imagery faces significant challenges due to the unique aerial perspective. A major bottleneck is the weak feature representation of small objects, which limits both detection accuracy and computational efficiency. To address this issue, we propose a Lightweight Detail-Sensitive Network (LDSNet). Specifically, LDSNet consists of three key components: (1) Lightweight Detail-Sensitive Downsampling (LDSDown), which combines anti-aliasing smoothing with dual-path feature extraction to preserve the spatial details of small objects during downsampling; (2) Shared Recursive Dilated Convolution (SRDC), which uses weight-shared multi-rate dilated convolutions to capture multi-scale context and enlarge the receptive field without introducing extra parameters; and (3) Deeply Decoupled Grouped Head (DGHead), which employs high-ratio grouped convolutions to significantly reduce the computational cost of processing high-resolution inputs. Extensive experiments on the VisDrone2019 and HIT-UAV datasets demonstrate that LDSNet achieves an excellent trade-off between accuracy and efficiency. Compared to the YOLOv11n baseline, LDSNet reduces parameters by 84.6% (from 2.6 M to 0.4 M) and FLOPs by 29.2% (from 6.5 G to 4.6 G), while improving mAP₅₀ by 2.2% on VisDrone2019 and achieving 94.5% on HIT-UAV.

Few-Shot Class-Incremental Learning (FSCIL) aims to sequentially learn new classes from very few labelled samples while preventing the forgetting of previously acquired knowledge, which has important practical value for remote sensing scene classification (RSSC). Recent studies have shown that applying a Vision Transformer (ViT) pre-trained on natural image datasets to FSCIL tasks can achieve significantly superior performance. Nevertheless, a substantial domain distribution gap exists between natural images and remote sensing images, which leads to severe performance degradation when such models are directly transferred to RSSC. To address the domain gap alongside FSCIL’s inherent stability–plasticity dilemma and overfitting under data scarcity, we propose a Dynamic Expansion Mixture-of-Experts with Pre-trained Vision Transformer (DEM-ViT) framework. Specifically, to alleviate the domain discrepancy, DEM-ViT incorporates an Adapter-Based Mixture-of-Experts (ABMoE) module, which captures the diverse visual patterns of remote sensing scenes through feature reconstruction in the representation space and collaborative learning among multiple experts. Furthermore, to address the stability–plasticity dilemma in FSCIL, we propose a Dynamic Expert Expansion (DEE) strategy, which progressively expands the model capacity along the incremental sessions. DEE provides sufficient space for learning new knowledge while mitigating the forgetting of previous knowledge. In addition, we propose a Semantic-Guided Feature Alignment (SGFA) method to reduce the risk of overfitting under data-scarce conditions. SGFA leverages textual information to construct robust text prototypes and uses them to calibrate the visual feature space. Extensive experiments across three benchmarks indicate that our framework exhibits highly competitive performance compared with state-of-the-art methods.