Jul 2026· Remote Sensing· Vol 18, pp. 2208· 0 citations
TL;DR
In the label noise suppression strategy (LNSS), a contrastive learning mechanism based on semantic consistency is introduced to constrain the aggregation of similar samples in the feature space, thereby reducing the adverse impact of noisy samples on model optimization.
Abstract
In aerial image small-object detection, complex imaging perspectives, arbitrary object orientations, and long-tailed category distributions jointly exacerbate sample imbalance, which significantly degrades detection stability and leads to frequent misclassification of minority categories. To address these challenges, this paper proposes a novel training framework termed SCUD. Specifically, in the label noise suppression strategy (LNSS), a contrastive learning mechanism based on semantic consistency is introduced to constrain the aggregation of similar samples in the feature space, thereby reducing the adverse impact of noisy samples on model optimization. In addition, a scale-aware resampling strategy (SARS) is designed to alleviate noise amplification and overfitting caused by excessive repetition of small objects during training. Furthermore, an adaptive instance selection mechanism (AISM) is developed by jointly modeling prediction uncertainty and global statistical priors, enabling the model to dynamically emphasize learning from informative samples. Extensive experiments are conducted on two publicly available unmanned aerial vehicle (UAV) aerial image datasets to validate the effectiveness of the proposed approach. The proposed method achieves an mAP50 of 70.7% on the DOTA-v1.0 dataset and 88.1% on the DIOR dataset. Notably, the detection accuracy of several rare categories is significantly improved, further demonstrating the effectiveness of the proposed method in addressing sample imbalance in aerial image small-object detection.
In Deep Learning Remote Sensing, data quantity is rarely the limiting factor. A single high-resolution satellite image can yield thousands of training patches. What determines model performance, yet remains largely overlooked, is the quality of those patches. To date, the choice of sampling method has rarely been treated as a methodological decision. Conventional approaches, namely sliding-window and random sampling, introduce two compounding data-quality problems: severe class imbalance caused by the overproduction of background-only patches and negative learning arising from incomplete annotations, where unlabeled objects are implicitly treated as negative examples during training. To address these limitations at the data construction stage, we propose object-centric patch sampling, a model-independent strategy that anchors each training patch to the geometric centroid of an annotated object. This design ensures that every object-anchored patch contains at least one target instance and substantially reduces exposure to unlabeled regions that generate false-negative supervision signals; only a small, deliberately controlled proportion of background-only patches is retained to preserve contextual variety without reinstating background dominance. The method is evaluated on three heterogeneous remote sensing datasets spanning satellite (Sentinel-2, 10 m), aerial (NAIP, 1 m), and UAV (0.25 m) imagery, covering cotton field segmentation, rural building extraction, and water body delineation, respectively. Using a U-Net architecture under identical training conditions, the proposed approach achieves IoU scores of 0.929, 0.896, and 0.912 on the three datasets, respectively, outperforming sliding-window sampling by up to 19.6 percentage points in IoU and consistently delivering higher F1-scores across all experimental configurations. Evaluation under DeepLabV3+ gives a mean IoU of 0.9504 and a mean F1-score of 0.9772 in a multi-class segmentation task, indicating that the gains are not specific to a single architecture. Unlike model-level solutions such as focal loss or class reweighting, the proposed method improves training data quality at its source and integrates seamlessly into any deep learning pipeline without architectural modifications.
A context-gated dynamic perception framework that treats small-object feature degradation as a coupled problem of representation, fusion, and prediction and indicates a practical accuracy-efficiency trade-off for dense aerial small-object perception.
Guangjun Gao, Ruibing Xie· Pattern Analysis and Applica...· 0 citations
Tiny object detection in remote sensing typically faces the challenges of being submerged in backgrounds, limited feature representation, and high sensitivity to prediction errors due to the small size and diverse shapes. To address these challenges, a geometric guided noise reduction super-resolution (SR) network is proposed. First, a dual-branch denoising and SR feature pyramid network is proposed, which integrates an adaptive dynamic noise reduction module and an inference decoupled auxiliary SR branch, while a progressive loss-annealing strategy is further introduced to reduce reliance on the SR branch during inference, meeting the requirements of lightweight and high-performance remote sensing tiny object detection. Second, a geometric characteristic regression metric is proposed, which comprehensively considers the location accuracy and the shape similarity between the prediction and ground truth boxes, thereby improving bounding-box quality and detection precision. Extensive experiments have been conducted on the remote sensing tiny object datasets AI-TOD v1, AI-TOD v2, USOD, and VisDrone. Specifically, it reaches an AP of 31.6 on AI-TOD v1, 30.5 on AI-TOD v2, 37.4 on USOD, and 30.5 on the VisDrone, demonstrating its capability for tiny object detection in remote sensing.
Ming-Xue Yang, He Chen, Ning Zhang et al.· IEEE Journal of Selected Top...· 0 citations
The proposed framework features a redesigned cross-scale feature fusion module, CCFM-S2, which utilizes the SPD-Conv operator for information preserving downsampling and explicitly integrates high-resolution shallow features (S2 layer), thereby infusing indispensable spatial details into the feature hierarchy for small targets.
Few-Shot Synthetic Aperture Radar (SAR) object detection aims to identify and localize unseen categories using only a small number of annotated support samples. However, performance is fundamentally constrained by two coupled challenges: speckle noise and structured clutter cause unstable region proposals and support–query mismatch, and extreme data scarcity leads to prototype drift and miscalibrated decision boundaries for novel classes. To tackle these issues in a coordinated framework, we propose Prior Class Perceptron Network (PCPNet), a support-conditioned detection framework that jointly improves proposal generation, the Prior Class Perceptron (PCP). First, the Adaptive Dense Proposal Module (ADPM) is designed to replace hard region selection with soft region-aware masks and uncertainty-aware mixture perception, enabling tighter, more reliable proposals in cluttered SAR scenes. Second, the PCP is utilized to decode prompted class perceptron from sparse supports, together with a self-calibrated prototype refinement strategy to mitigate prototype drift and stabilize novel scoring. Finally, by incorporating shot-dependent and uncertainty-aware adaptive margins into the Cross-Entropy (CE) loss, the PCPNet model improves inter-class separation and intra-class compactness under limited supervision. Extensive experiments on SAR-AIRCRAFT-1.0 and MSTAR-FEWSHOT demonstrate that PCPNet consistently achieves SOTA performance across three novel splits and multiple shot settings. In particular, PCPNet attains 27.47% mAP on SAR-AIRCRAFT-1.0 and 52.86% mAP on MSTAR-FEWSHOT under the single-run protocol, while also exhibiting strong robustness and cross-dataset generalization. Additional diagnostic, sensitivity, and qualitative analyses further verify the effectiveness and robustness of the proposed framework. These results indicate that jointly modeling uncertainty, prior class knowledge, and shot-dependent decision calibration provides an effective solution for few-shot SAR object detection.
Shengjie Lei, Ziqi Wu, Xinyi Cai et al.· Remote Sensing· 0 citations
Object detection in unmanned aerial vehicle (UAV) imagery presents unique challenges, including extreme scale variations, highly imbalanced spatial distribution of objects, and computational constraints inherent to drone platforms. To address these issues, this paper introduces a novel self-regularized object detection framework specifically designed for UAV perspectives. The proposed method adopts LSNet as the backbone network, guided by the "See Large, Focus Small" principle, to enhance feature extraction capability for objects of varying scales under wide-angle UAV views. The framework incorporates two key components: a Self-Regularized Temporal Ratio Encoder (SRTR Encoder) that stabilizes training and improves bounding box quality by leveraging temporal aspect ratio constraints, and an Entropy Constrained Sampler (EC Sampler) that selects informative hard negative samples based on information entropy, thereby suppressing less relevant background regions and increasing training efficiency. Extensive experiments conducted on the VisDrone benchmark demonstrate that the proposed approach significantly outperforms existing methods, achieving a notable improvement in mean Average Precision (mAP) while simultaneously reducing computational costs. For instance, when integrated with the CEASC detector, the framework increases mAP from 25.4% to 28.0% while reducing GFLOPs from 105.90 to 92.61. The results validate the effectiveness of our method in addressing the specific challenges of UAV-based object detection, offering a balanced solution for accuracy and efficiency in resource-constrained scenarios.
Chenguang Zhang, Yangming Guo, Jian-Long Yu et al.· International Conference on...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
