TL;DR
A lightweight object detection framework, termed MDCF-YOLO, which achieves a superior accuracy-efficiency trade-off compared to state-of-the-art lightweight object detectors and exhibits similarly competitive performance on the AI-TOD dataset, further validating its effectiveness and generalization capability in UAV remote sensing scenarios.
(1) Objective: Remote sensing object detection faces significant challenges, including complex background interference, large variations in target scales, and insufficient multi-scale feature representation, which often result in missed detections of small objects, inaccurate localization, and inadequate feature fusion. (2) Methods: To address these issues, this paper proposes a Multi-scale Attention Enhancement Feature Fusion Network (MAEF-Net) based on YOLOv12. The proposed method systematically optimizes the overall detection pipeline from three aspects, namely feature enhancement, feature fusion, and prediction refinement. Specifically, it enhances local texture representation and high-level semantic information in the backbone, improves multi-scale feature interaction during the feature fusion stage, and adaptively filters fused features in the detection stage to strengthen target responses while suppressing background noise, thereby improving the detection performance for complex scenes, small objects, and densely distributed targets. (3) Results: To validate the effectiveness of the proposed method, extensive experiments are conducted on two public remote sensing object detection datasets, namely RSOD and NWPU VHR-10. Experimental results demonstrate that, compared with the baseline YOLOv12, MAEF-Net improves the mAP by 1.86% and 1.05% on the RSOD and NWPU VHR-10 datasets, respectively. Furthermore, compared with the latest YOLOv13, the proposed method achieves additional mAP improvements of 1.53% and 0.43%, respectively. Moreover, MAEF-Net achieves a favorable balance between detection accuracy and computational efficiency while maintaining relatively low computational complexity and high inference speed. Ablation studies, comparative experiments, and qualitative visualization further demonstrate the effectiveness, robustness, and generalization capability of the proposed method in complex remote sensing scenarios.
An Adaptive and Scalable YOLO model named AS-YOLOR (Adaptive and Scalable YOLO for Rotated object detection), based on the YOLOv8 baseline is proposed, providing a solution with strong practical potential for achieving efficient and high-precision detection of small, rotated objects.
Aerial small object detection is confronted with critical challenges including extreme scale variation, dense target distribution, heavy background interference and strict real-time deployment constraints on unmanned aerial vehicle platforms. Traditional real-time detectors fail to balance detection accuracy and computational efficiency in such scenarios, suffering from severe cross-layer feature misalignment and small object feature attenuation caused by the mismatch between deep semantic features and shallow spatial features. To address these issues, this paper proposes a lightweight real-time detection framework for aerial dense small objects, named PHAF-YOLO. Three targeted optimizations are designed: the Progressive Multi-Kernel Enhancement Unit is embedded in the backbone to expand the effective receptive field with low overhead and alleviate deep feature attenuation via progressive multi-kernel convolution; the Hierarchical Dual-stage Adaptive Fusion Block is applied in the neck to dynamically screen multi-scale features and suppress fusion redundancy through dual-stage adaptive fusion, improving cross-scale information utilization for dense small objects; the Target-Centric Soft Regression Loss combines object-centered geometric constraints with soft target assignment to mitigate gradient instability from ambiguous positive-negative sample boundaries and boost high-IoU localization accuracy. Extensive experiments on VisDrone2019, UAVDT2018 and AI-TOD2021 demonstrate that PHAF-YOLO achieves a superior accuracy-efficiency trade-off across all model scales and outperforms mainstream real-time detectors. Ablation and validation experiments further verify the independent effectiveness and synergistic effect of all core modules. The source code of this work is publicly available at: https://github.com/csy001x/PHAF-YOLO.
Aiming at the challenges in small object detection in aerial images, such as extremely small object scales, weakened features, and background interference, this paper proposes a layered enhancement network named LE-YOLO. The proposed framework initiates enhancement at the feature level by integrating the DySample module, which replaces standard upsampling operators through a dynamic point sampling mechanism. This design significantly improves the reconstruction of fine-grained details with negligible computational overhead. Building upon this improved feature representation, a cross-layer feature fusion module is incorporated to aggregate multi-level features, effectively amplifying semantic coherence in regions of interest while suppressing extraneous background noise. Further refining the detection capability, the model introduces a layered enhancement head architecture. Within this head, a novel semantic information enhancement module employs multi-scale dilated convolutional layers to accommodate the substantial size variation characteristic of small aerial objects. The regression branch is subsequently optimized through receptive field adaptive convolutions, enhancing adaptability to diverse object scales. Complementing these architectural innovations, the training process utilizes the normalized Gaussian Wasserstein distance loss to diminish sensitivity to minor localization inaccuracies, particularly beneficial for small targets. Results demonstrate that compared to YOLOv8s, LE-YOLO improves AP50 by 3.1% and 1.7% on the AITOD and SIMD datasets, and outperforms other advanced models in comparative tests.
To address the challenges of small scales, weak features, complex backgrounds, and misalignment in multi-scale fusion for small-object detection in remote sensing images, this study proposes a frequency-guided fusion reconstruction YOLO (FFR-YOLO), an improved YOLOv8 framework. The method performs joint optimization across three levels: the backbone, neck, and front end of the detector head. In the backbone, a frequency-guided anti-alias progressive downsampling module utilizes Haar wavelet decomposition to replace traditional strided convolutions and incorporates a low-frequency-guided high-frequency gating mechanism to mitigate detail loss and background noise interference during downsampling. In the neck, a bridge-guided bidirectional reconstruction fusion module (BRFM) enhances the collaborative reconstruction of multi-scale semantic and detailed information via multi-source weighted fusion and cross-path bridging interactions. At the front end of the detector head, a recalibrated dual-branch local–global fusion (RDLGF) module implements dynamic allocation and complementary fusion of dual-path features. Experiments were conducted on two datasets, DIOR and NWPU VHR-10. The results demonstrate that FFR-YOLO achieves a mAP@0.5 of 85.8% and a mAP@0.5:0.95 of 63.1% on DIOR and 93.6% and 62.1% on NWPU VHR-10. These outcomes present improvements over the baseline YOLOv8, validating the effectiveness and practical value of the proposed method for small-object detection in remote sensing scenarios.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
