In order to solve the problems of high leakage rate, high false detection rate, low detection success rate and large model volume of small targets in the traditional target detection algorithm for Unmanned Aerial Vehicle (UAV) aerial images, a lightweight You Only Look Once (YOLO) v8 algorithm model Lightweight (LW)-YOLO v8 is proposed. By increasing the channel attention mechanism Squeeze-and-Excitation (SE) module, this method can adaptively improves the model’s ability to extract features from small targets; at the same time, the lightweight convolution technology is introduced into the Conv module, where the ordinary convolution is replaced by the GSConv module, which can effectively reduce the model computational volume; on the basis of the GSConv module, a single aggregation module VoV-GSCSPC is designed to optimize the model structure in order to achieve a higher computational cost-effectiveness. The experimental results show that the LW-YOLO v8 model’s mAP@0.5 metrics on the VisDrone2019 dataset are more favorable than those on the YOLO v8n model, improving by 3.8 percentage points, and the computational amount is reduced to 7.2 GFLOPs. The LW-YOLO v8 model proposed in this work can effectively accomplish the task of detecting small targets in aerial images from UAV at a lower cost.

Addressing the challenges in detecting surface floating litter in artificial lakes, including complex environments, uneven illumination, and susceptibility to noise and weather, this paper proposes an efficient and lightweight Ghost-YOLO (You Only Look Once) v8 algorithm. The algorithm integrates advanced attention mechanisms and a small-target detection head to significantly enhance detection performance and efficiency. Firstly, an SE (Squeeze-and-Excitation) mechanism is incorporated into the backbone network to fortify the extraction of resilient features and precise target localization. This mechanism models feature channel dependencies, enabling adaptive adjustment of channel importance, thereby improving recognition of floating litter targets. Secondly, a 160 × 160 small-target detection layer is designed in the feature fusion neck to mitigate semantic information loss due to varying target scales. This design enhances the fusion of deep and shallow semantic information, improving small target feature representation and enabling better capture and identification of tiny floating litter. Thirdly, to balance performance and efficiency, the GhostConv module replaces part of the conventional convolutions in the feature fusion neck. Additionally, a novel C2fGhost (CSPDarknet53 to 2-Stage Feature Pyramid Networks Ghost) module is introduced to further reduce network parameters. Lastly, to address the challenge of occlusion, a new loss function, WIoU (Wise Intersection over Union) v3 incorporating a flexible and non-monotonic concentration approach, is adopted to improve detection rates for surface floating litter. The outcomes of the experiments demonstrate that the Ghost-YOLO v8 model proposed in this paper performs well in the dataset Marine, significantly enhances precision and recall by 3.3 and 7.6 percentage points, respectively, in contrast with the base model, mAP@0.5 and mAP@0.5:0.95 improve by 5.3 and 4.4 percentage points and reduces the computational volume by 1.88 MB, the FPS value hardly decreases, and the efficient real-time identification of floating debris on the water’s surface can be achieved cost-effectively.

Location information on Twitter plays a critical role in emergency detection, event recommendation, and disaster warning. However, only a limited amount of Twitter data are geotagged. Previous research has presented various models for inferring location based on text, social relations, and contextual data, yielding highly promising results. Nonetheless, these existing methods have certain limitations that need to be addressed. Firstly, most of the existing methods overlook the role of local celebrities (well-known users in the local community) as indicators of location within the social network. Secondly, they fail to consider the associations between words in tweets, resulting in insufficiently rich features extracted from the tweets. We propose a multi-view-based location inference model called MVGeo to overcome these limitations. In the network view, our approach employs the Gaussian Mixture Model (GMM) to identify and retain local celebrities, thereby strengthening user location associations. In the tweet view, we construct a heterogeneous graph based on the co-occurrence relationship between words in tweets and the user’s mentioned relationship with the words. This allows us to fully leverage the local correlation between words and the global correlation to extract tweet features more comprehensively. Finally, we employ a modified multi-layer graph convolutional network, called Gate-GCN, to fuse the network and tweet information. This expansion of the feature space enables us to extract sample features from multiple perspectives. To demonstrate the effectiveness of MVGeo, we conduct exhaustive experimental evaluations on a publicly available dataset and compare its performance against several state-of-the-art benchmark models. The results confirm the superior performance of the proposed model.