Melanoma is a malignant tumor originating from the lesions of skin cells. Medical image segmentation tasks for skin lesion play a crucial role in quantitative analysis. Achieving precise and efficient segmentation remains a significant challenge for medical practitioners. Hence, a skin lesion segmentation model named MSDUNet, which incorporates multi-scale deformable block (MSD Block) and dual-input dynamic enhancement module(D2M), is proposed. Firstly, the model employs a hybrid architecture encoder that better integrates global and local features. Secondly, to better utilize macroscopic and microscopic multiscale information, improvements are made to skip connection and decoder block, introducing D2M and MSD Block. The D2M leverages large kernel dilated convolution to draw out attention bias matrix on the decoder features, supplementing and enhancing the semantic features of the decoder’s lower layers transmitted through skip connection features, thereby compensating semantic gaps. The MSD Block uses channel-wise split and deformable convolutions with varying receptive fields to better extract and integrate multi-scale information while controlling the model’s size, enabling the decoder to focus more on task-relevant regions and edge details. MSDUNet attains outstanding performance with Dice scores of 93.08% and 91.68% on the ISIC-2016 and ISIC-2018 datasets, respectively. Furthermore, experiments on the HAM10000 dataset demonstrate its superior performance with a Dice score of 95.40%. External validation experiments based on the ISIC-2016, ISIC-2018, and HAM10000 experimental weights on the PH2 dataset yield Dice scores of 92.67%, 92.31%, and 93.46%, respectively, showcasing the exceptional generalization capability of MSDUNet. Our code implementation is publicly available at the Github.

Natural scene text detection has attracted much attention in the research field of computer vision, and it has been widely used in many applications, i.e., unmanned driving, robot sensing. Some methods have been proposed for horizontal and oriented text detection, but detecting irregular shapes and highly varying orientated text is still a challenging problem. To tackle this problem, we propose a robust arbitrary shape text detection method called Soft Dilated network (SDnet). The proposed method has two essential steps: (1) feature extraction by backbone; (2) post-processing approach to generate elaborated polygons or boundaries. In particular, the backbone is based on soft attention mechanism and dilated convolution. The soft attention mechanism learns and obtains importance feature from each feature channel, and dilated convolution can effectively aggregate multi-scale contextual information without losing the resolution, and enhance the robust of the network model. The proposed method can accurately detect curve text and discriminate text and non-text areas in an efficient fashion. In addition, Jaccard coefficient is used as loss function to promote the post-processing capability of detecting sparse-arranged and arbitrary shape text. Based on the aforementioned technique, the proposed method an effectively handle the problem of sparse arranged arbitrary natural scene text detection. Experiments were conducted on three benchmark datasets: curved text dataset CTW1500, Total-Text and oriented dataset ICDAR2015, and the results show that when compared with the state-of-the-art text detection methods, the proposed method is more robust and it can find smaller text blocks in the image due to the Loss Function calculation with Jaccard coefficient. Furthermore, we performed multiple sets of ablation experiments, verify the effectiveness of the propose method.

With the growing use of drones, efficient detection algorithms are crucial, especially under adverse weather conditions. Most existing drone detection algorithms perform well only in clear weather, resulting in significant performance drops in foggy conditions. This study focuses on improving drone detection in foggy environments using the Mean Teacher framework for domain adaptation. The Mean Teacher framework’s performance relies on the quality of the teacher model’s pseudo-labels. To enhance the quality of the pseudo-labels from the teacher model, we introduce Foggy Drone Teacher (FDT), which includes three key components: (1) Adaptive Style and Context Augmentation to reduce domain shift and improve pseudo-label quality; (2) Simplified Domain Alignment with a novel adversarial strategy to boost domain adaptation; and (3) Progressive Domain Adaptation Training, a two-stage process that helps the teacher model produce more stable and accurate pseudo-labels. In addition, owing to the lack of publicly available data, we created Foggy Drone Dataset (FDD) to support this research. Extensive experiments show that our model achieves a 21.1-point increase in $A{P}_{0.5}$ compared to the baseline and outperforms state-of-the-art models. This method significantly improves drone detection accuracy in foggy conditions.

Semantic segmentation of remote sensing imagery is a pivotal task for intelligent interpretation, with critical applications in urban monitoring, resource management, and disaster assessment. Recent advancements in deep learning have significantly improved RS image segmentation, particularly through the use of convolutional neural networks, which demonstrate remarkable proficiency in local feature extraction. However, due to the inherent locality of convolutional operations, prevailing methodologies frequently encounter challenges in capturing long-range dependencies, thereby constraining their comprehensive semantic comprehension. Moreover, the preprocessing of high-resolution remote sensing images by dividing them into sub-images disrupts spatial continuity, further complicating the balance between local feature extraction and global context modeling. To address these limitations, we propose DMA-Net, a Dynamic Morphology-Aware Segmentation Network built on an encoder–decoder architecture. The proposed framework incorporates three primary parts: a Multi-Axis Vision Transformer (MaxViT) encoder achieves a balance between local feature extraction and global context modeling through multi-axis self-attention mechanisms; a Hierarchy Attention Decoder (HA-Decoder) enhanced with Hierarchy Convolutional Groups (HCG) for precise recovery of fine-grained spatial details; and a Channel and Spatial Attention Bridge (CSA-Bridge) to mitigate the encoder–decoder semantic gap while amplifying discriminative feature representations. Extensive experimentation has been conducted to demonstrate the state-of-the-art performance of DMA-Net, which has been shown to achieve 87.31% mIoU on Potsdam, 83.23% on Vaihingen, and 54.23% on LoveDA, thereby surpassing existing methods.

The nnU-Net framework effectively automates hyperparameter selection; however, its fixed internal configurations—notably convolution kernel sizes—restrict its flexibility. This limitation is pronounced in 3D medical imaging, where anatomical structures undergo continuous spatial evolution along the Z-axis. In this study, we introduce a self-adaptive convolution module designed to dynamically tune the effective receptive field, matching the dynamic structural transformations of organs. By employing a differentiable soft-attention mechanism to aggregate candidate kernels, the network adaptively optimizes its scale sensitivity. This integration allows MSA ² -Net to capture both global context and local nuances within feature maps. The module is strategically embedded into two core components: the multi-scale convolution bridge and the multi-scale amalgamation decoder. In the Bridge, it refines CSWin Transformer outputs by aligning features with the inherent spatial continuity of volumetric data, thereby mitigating redundancies that might otherwise hinder decoding. Simultaneously, the multi-scale amalgamation decoder leverages this module to precisely reconstruct organ details as their size and shape fluctuate across slices. This mechanism ensures the decoder preserves seamless topological intricacies within the feature maps, yielding superior segmentation accuracy. Leveraging this architecture, MSA ² -Net achieves competitive Dice scores of 86.49%, 92.56%, 93.37%, and 92.98% on the Synapse, ACDC, Kvasir, and ISIC2017 datasets, respectively. Extensive experiments validate the model's robustness in handling complex spatial variations across diverse medical modalities.