AS-YOLOR: An Improved YOLO Model for Small Object Detection in Aerial Images

SFC-YOLO: An Accuracy-Enhanced and Parameter-Efficient YOLO Framework for Small Vehicle Detection in Aerial Images

Small vehicle detection in aerial images is important for intelligent transportation, low-altitude inspection, urban monitoring, and vision-based sensing systems. Vehicles in aerial images often occupy few pixels and are affected by complex backgrounds, shadows, viewpoint changes, weak texture, and similar class appearances, which can cause missed detections and false alarms. To address these issues, this paper proposes SFC-YOLO, an accuracy-enhanced and parameter-efficient framework based on YOLO11n. A Feature Complementary Block (FCB) is placed at the P5/32 high-level feature stage to enhance local-detail and semantic-context compensation; a Dynamic Feature Alignment Upsampling unit (DFAU) is inserted into the first P5-to-P4 upsampling path to improve content-adaptive feature alignment; and a P5-only Hidden-State Attention (HSA) module is used in the final P5 detection branch to strengthen global semantic interaction. Experiments on the VEDAI eight-class vehicle dataset show that SFC-YOLO reduces the parameter count from 2.584 M to 2.466 M while improving the five-run mean Precision from 0.602 to 0.665, Recall from 0.599 to 0.608, and mAP50 from 0.614 to 0.652. Since the GFLOPs increase from 6.3 to 8.4, SFC-YOLO should be interpreted as a parameter-reduced but not FLOP-reduced framework. The main trade-off is improved detection accuracy with fewer trainable parameters at the cost of higher theoretical computation. Five repeated experiments yield an average mAP50 of 0.6523 ± 0.0117, quantifying the run-to-run variation under the current training protocol.

FR-YOLO: A Focus-and-Reconstruct Mechanism for Lightweight Small Object Detection in Drone Imagery

Automatic detection of small objects in UAV imagery is a challenging problem that is of great interest in aerial surveillance and intelligent transport. The detection model must cope well with feature degradation of tiny targets, drastic scale variations, and strict constraints on onboard com-putational resources. This paper proposes a lightweight attention-based network, called FR-YOLO, to address the "focus" and "reconstruct" chal-lenges in small object detection. We introduce two novel components: the Local Feature Enhancement (LFE) module to precisely suppress back-ground noise via spatial attention, and the Content-aware Feature Reassem-bly (CFR) module to rectify spatial feature misalignment and recover fine-grained details during upsampling. The proposed modules are seamlessly integrated into the YOLO11n backbone. Extensive experiments on the au-thoritative VisDrone2019 benchmark indicate that FR-YOLO outperforms the YOLO11n baseline by a significant margin of 1.5% and surpasses the widely adopted YOLOv8n by 0.7% (achieving 36.4% mAP), all while maintaining a 16% smaller model size (2.53M parameters) compared to YOLOv8n. We have made the code available to the public at: https://github.com/Liu999hongwei/FR-YOLO.

DAGE-YOLO: improved YOLO for oriented object detection in remote sensing imagery

Remote sensing images are often featured by small object sizes, varying rotation angles, and background interference, which makes oriented object detection (OOD) from these images challenging. Here, an improved YOLO11s model, termed DAGE-YOLO, is proposed to resolve this challenge. Specifically, parallelized patch-aware attention (PPA) and efficient multi-scale attention (EMA) are incorporated into the backbone of the original YOLO11s to preserve features of small objects and improve cross-scale context modeling. Next, the conventional loss is displaced by the Gaussian Wasserstein Distance (GWD) loss. Moreover, an oriented bounding box (OBB) detection head is introduced to allow the model to achieve precise oriented box localization. Experiments on the DOTAv1.5 and HRSC2016 datasets show that DAGEYOLO achieves an mAP50 8.8% and 7.1% higher than that of the baseline model on the two datasets, respectively, and outperforms several mainstream YOLO-based OOD models. Though this DAGE-YOLO model introduces a moderate increase in the computing overhead, it achieves strong detection accuracy and robustness, offering an effective solution for complex OOD tasks in remote sensing.

SPAF: Enhancing YOLO-Based Object Detection Using Sparse Attention and Progressive Adaptive Fusion

Detecting objects in Autonomous Aerial Vehicles (AAVs) imagery remains a challenging task due to large variations in object sizes and complex backgrounds, which often leads to missed detections and reduced accuracy, especially for small objects that contain limited visual information and are easily lost during feature extraction. To address these challenges, this paper proposes SPAF, an enhanced YOLO-based model built upon YOLO26 that improves object detection through the integration of a multi-scale attention mechanism, progressive feature aggregation, and adaptive feature fusion. The proposed architecture enhances the backbone network by integrating a Sparse Vision Attention (SVA) module to facilitate the extraction of fine-grained and task-relevant features in complex aerial scenes. In the neck, the Progressive Bidirectional Auxiliary Pyramid Network (PBAP-Net) facilitates efficient cross-scale feature aggregation by combining top-down semantic information with bottom-up spatial details, thereby strengthening multi-scale representation. Finally, Adaptive Channel Multi-Scale Fusion Head(ACMF-Head) refines the final feature representations by adaptively fusing multi-scale information and reducing inconsistencies across feature levels, resulting in more reliable and robust predictions. Experimental results on the VisDrone2019 dataset demonstrate that SPAF achieves 46.8% mAP@0.5, which is 7.5% higher compared to the baseline model. These results confirm the effectiveness of the proposed approach for robust AAV object detection in challenging aerial environments.

QS-YOLOv8: a lightweight and efficient detector for small objects

HASO-DETR: hybrid attention small object detection based on RT-DETR

In unconstrained environments such as aerial and surveillance imagery, achieving precise detection of small objects continues to pose significant challenges in computer vision because of constrained pixel data and cluttered backgrounds, and high sensitivity to bounding box localization errors. Although existing detectors have made progress, they often suffer from the gradual loss of fine-grained features in deep networks and inefficient multiscale feature fusion. Motivated by these challenges, this study designs the HASO-DETR small object detection algorithm.The proposed framework features three key contributions: 1 The Hybrid Dilated Residual Attention Block (HDRAB) embedded in the backbone network, which employs multi-scale dilated convolutions and channel attention mechanisms to enhance and preserve critical features of small objects during downsampling stages; 2 A redesigned cross-scale feature fusion module, CCFM-S2, which utilizes the SPD-Conv operator for informationpreserving downsampling and explicitly integrates high-resolution shallow features (S2 layer), thereby infusing indispensable spatial details into the feature hierarchy for small targets; 3 The adoption of Employing the NormalizedWasserstein Distance (NWD) as a regression loss, we model bounding boxes as Gaussian distributions to effectively reduce sensitivity to minor deviations in bounding box coordinates. Extensive experiments on challenging benchmark datasets VisDrone2019 , demonstrate that HASO-DETR sets a new state-of-the-art performance, outperforming numerous mainstream detectors.