Landslides are a type of sudden and highly destructive geological hazard. Traditional detection methods often suffer from delayed response and low efficiency. In recent years, deep learning-based object detection techniques have attracted increasing attention in disaster recognition tasks, particularly transformer-based detection models, which exhibit significant advantages in global feature modeling. However, landslide targets in remote sensing imagery often present challenges such as large-scale variation, blurred boundaries, and texture interference. To address these issues, this study proposes an improved detection algorithm based on the RT-DETR-r18 framework by integrating multiple specialized modules. First, the DDC3 module is designed to enhance the recognition of fine boundaries and local textures, thereby improving the feature extraction capacity of the backbone network. Second, an Efficient Additive Attention (EAA) mechanism is introduced to suppress redundant information and strengthen the model's focus on critical regions, improving detection precision. Finally, the CGAFusion module is employed, which utilizes a triple-attention strategy to collaboratively regulate feature weights. This module enhances the model’s ability to filter salient features while preserving global contextual information, leading to more accurate landslide edge detection. A dual-class dataset comprising landslides and storms is constructed from multi-source imagery for evaluation. The experimental results show that the proposed method outperforms existing models in several dimensions including mAP@0.5 and F1 score, demonstrating strong detection accuracy. Code is available at https://github.com/sanyauChenCoder/Landslide_02.git.

Early detection and accurate diagnosis of brain tumors are crucial for patient treatment and prognosis. Although traditional medical image analysis methods have made certain progress, they still face challenges such as relying on manual experience and insufficient accuracy. To address this, this paper proposes a brain tumor detection method based on deep learning, which combines an improved YOLOv8n network, an innovative C2f-DCNv2 module, and the DySample upsampling operator. By introducing the C2f-DCNv2 module, the model integrates deformable convolutions and multi-branch structures in the convolutional neural network (CNN), effectively enhancing the ability to extract tumor features. The DySample upsampling operator dynamically adjusts the sampling grid, significantly improving image detail restoration. In experiments, the proposed model achieved significant improvements over YOLOv8n. The recall rate increased from 86.5% in YOLOv8n to 88.2% in the proposed model, a 1.7% improvement. The mAP@0.5 (mean average precision) reached 93.0% in the proposed model, compared to 92.2% in YOLOv8n, a 0.8% improvement. The mAP@0.5-0.95 reached 66.5%, compared to YOLOv8n's 64.9%, an improvement of 1.6%. These results demonstrate that the proposed method outperforms existing YOLO series models in multiple key metrics, offering stronger detection accuracy and robustness. The method provides an efficient and accurate solution for automatic brain tumor detection and lays a foundation for future applications of deep learning in medical image analysis.

Small object detection has always been a challenging task in the field of remote sensing image detection. Due to the small proportion and limited pixel size of small objects in images, especially when using Convolutional Neural Networks (CNNs), the downsampling operation in traditional algorithms often leads to the loss of detailed information of small objects, causing missed detection issues. To address this problem, this paper proposes an improved YOLOv8 algorithm. We designed an adaptive feature extraction and multi-scale fusion module, which enhances the expressive ability of features and effectively extracts the detailed information of small objects. We incorporated the AFGCAttention attention mechanism to strengthen the network's focus on key regions, suppress irrelevant background information, and improve the model's ability to recognize small objects. To overcome the resolution loss problem in small object detection, we adopted the CARAFE (Content-Aware ReAssembly of FEatures) upsampling operator. By reorganizing feature maps with content-awareness, it avoids the blurriness and information loss commonly found in traditional upsampling methods, especially showing significant advantages in the reconstruction of small object details, making their boundaries clearer and more accurate. Meanwhile, to improve the accuracy of bounding box regression, we combined the GIoU loss function to optimize the geometric shape matching between the target and predicted boxes, solving the issue of inaccurate bounding box localization in small object detection and improving localization precision. Experimental results show that the proposed algorithm achieves significant accuracy improvement in small object detection tasks, maintaining high detection robustness in complex background scenes. The improved model reaches a mean average precision (mAP) of 83.0% and an accuracy of 85.0%, which is an increase of 2.0% and 5.0%, respectively, compared to the baseline model. Compared with existing methods, this approach has significant advantages in detection accuracy, localization precision, and model computational efficiency, especially demonstrating outstanding performance in small object detection.