Local Contextual Information Research Articles

A multi-scale contextual attention network for remote sensing visual question answering

Remote sensing visual question answering (RSVQA) is a user-friendly method used for analyzing remote sensing images (RSIs) in various tasks. However, current methods often overlook geospatial objects, which possess a multi-scale representation and require contextual information. Furthermore, limited research has been conducted on modeling and reasoning the long-distance dependencies between entities, resulting in one-sided and inaccurate answer predictions. To overcome these limitations, we propose the Scale-Aware Multi-level Feature Pyramid Network (SAMFPN), which integrates contextual and multi-scale information using a Feature Pyramid Network (FPN) and Co-Attention mechanisms. The SAMFPN module incorporates a multi-level FPN to capture both global and local contextual information. Additionally, it introduces a Visual-Question Collaboration Fusion (VQCF) module that simultaneously embeds and learns visual and textual information. Our experimental results demonstrate the superior accuracy and robustness of our proposed model compared to existing models. These outcomes indicate that SAMFPN effectively captures multi-scale contextual information, making it a reliable solution for RSVQA tasks.

A LLM-Based Hybrid-Transformer Diagnosis System in Healthcare.

The application of computer vision-powered large language models (LLMs) for medical image diagnosis has significantly advanced healthcare systems. Recent progress in developing symmetrical architectures has greatly impacted various medical imaging tasks. While CNNs or RNNs have demonstrated excellent performance, these architectures often face notable limitations of substantial losses in detailed information, such as requiring to capture global semantic information effectively and relying heavily on deep encoders and aggressive downsampling. This paper introduces a novel LLM-based Hybrid-Transformer Network (HybridTransNet) designed to encode tokenized Big Data patches with the transformer mechanism, which elegantly embeds multimodal data of varying sizes as token sequence inputs of LLMS. Subsequently, the network performs both inter-scale and intra-scale self-attention, processing data features through a transformer-based symmetric architecture with a refining module, which facilitates accurately recovering both local and global context information. Additionally, the output is refined using a novel fuzzy selector. Compared to other existing methods on two distinct datasets, the experimental findings and formal assessment demonstrate that our LLM-based HybridTransNet provides superior performance for brain tumor diagnosis in healthcare informatics.

DE-Net: A Dual-Encoder Network for Local and Long-Distance Context Information Extraction in Semantic Segmentation of Large-Scale Scene Point Clouds

DE-Net: A Dual-Encoder Network for Local and Long-Distance Context Information Extraction in Semantic Segmentation of Large-Scale Scene Point Clouds

Multi-scale local-global transformer with contrastive learning for biomarkers segmentation in retinal OCT images

Quantitative analysis of biomarkers in Optical Coherence Tomography (OCT) images plays an import role in the diagnosis and treatment of retinal diseases. However, biomarker segmentation in retinal OCT images is very hard due to the large variations in size and shape of retinal biomarkers, blurred boundaries, low contrast, and speckle interference. We proposed a novel Multi-scale Local-Global Transformer network (MsLGT-Net) for biomarker segmentation in retinal OCT images. The network combines the proposed Multi-scale Fusion Attention (MFA) module, Local-Global Transformer (LGT) module, and Contrastive Learning Enhancement (CLE) module to tackle the challenges of biomarker segmentation. Specifically, the proposed MFA module aims to enhance the network’s ability to learn multi-scale features of retinal biomarkers by effectively combining the local detail information and contextual semantic information of biomarkers at different scales, and improve the representation ability for different classes of biomarkers. The LGT module is designed to learn local and global information adaptively from multi-scale fused features to address the challenge of small biomarker segmentation. In addition, to distinguish features between different types of retinal biomarkers, we propose the CLE module to enhance the feature representation of different biomarkers. Our proposed method is validated on one public dataset and one local dataset. The experimental results show that the proposed method is more effective than other state-of-the-art methods.

Real-Time Semantic Segmentation via a Densely Aggregated Bilateral Network.

With the growing demands of applications on online devices, the speed-accuracy trade-off is critical in the semantic segmentation system. Recently, the bilateral segmentation network has shown promising capacity to achieve the balance between favorable accuracy and fast speed, and has become the mainstream backbone in real-time semantic segmentation. Segmentation of target objects relies on high-level semantics, whereas it requires detailed low-level features to model specific local patterns for accurate location. However, the lightweight backbone of bilateral architecture limits the extraction of semantic context and spatial details. And the late fusion of the bilateral streams incurs the insufficient aggregation of semantic context and spatial details. In this article, we propose a densely aggregated bilateral network (DAB-Net) for real-time semantic segmentation. In the context path, a patchwise context enhancement (PCE) module is proposed to efficiently capture the local semantic contextual information from spatialwise and channelwise, respectively. Meanwhile, a context-guided spatial path (CGSP) is designed to exploit more spatial information by encoding finer details from the raw image and the transition from the context path. Finally, with multiple interactions between bilateral branches, the intertwined outputs from bilateral streams are combined in a unified decoder for a final interaction to further enhance the feature representation, which generates the final segmentation prediction. Experimental results on three public benchmarks demonstrate that our proposed method achieves higher accuracy with a limited decay in speed, which performs favorably against state-of-the-art real-time approaches and runs at 31.1 frames/s (FPS) on the high resolution of [Formula: see text] . The source code is released at https://github.com/isyangshu/DABNet.

DDoS‐MSCT: A DDoS Attack Detection Method Based on Multiscale Convolution and Transformer

Distributed denial‐of‐service (DDoS) attacks pose a significant threat to network security due to their widespread impact and detrimental consequences. Currently, deep learning methods are widely applied in DDoS anomaly traffic detection. However, they often lack the ability to collectively model both local and global traffic features, which presents challenges in improving performance. In order to provide an effective method for detecting abnormal traffic, this paper proposes a novel network architecture called DDoS‐MSCT, which combines a multiscale convolutional neural network and transformer. The DDoS‐MSCT architecture introduces the DDoS‐MSCT block, which consists of a local feature extraction module (LFEM) and a global feature extraction module (GFEM). The LFEM employs convolutional kernels of different sizes, accompanied by dilated convolutions, with the aim of enhancing the receptive field and capturing multiscale features simultaneously. On the other hand, the GFEM is utilized to capture long‐range dependencies for attending to global features. Furthermore, with the increase in network depth, DDoS‐MSCT facilitates the integration of multiscale local and global contextual information of traffic features, thereby improving detection performance. Our experiments are conducted on the CIC‐DDoS2019 dataset, and also the CIC‐IDS2017 dataset, which is introduced as a supplement to address the issue of sample imbalance. Experimental results on the hybrid dataset show that DDoS‐MSCT achieves accuracy, recall, F1 score, and precision of 99.94%, 99.95%, 99.95%, and 99.97%, respectively. Compared to the state of the art methods, the DDoS‐MSCT model achieves a good performance for detecting the DDoS attack to provide the protecting ability for network security.

Swin-GA-RF: genetic algorithm-based Swin Transformer and random forest for enhancing cervical cancer classification.

Cervical cancer is a prevalent and concerning disease affecting women, with increasing incidence and mortality rates. Early detection plays a crucial role in improving outcomes. Recent advancements in computer vision, particularly the Swin transformer, have shown promising performance in image classification tasks, rivaling or surpassing traditional convolutional neural networks (CNNs). The Swin transformer adopts a hierarchical and efficient approach using shifted windows, enabling the capture of both local and global contextual information in images. In this paper, we propose a novel approach called Swin-GA-RF to enhance the classification performance of cervical cells in Pap smear images. Swin-GA-RF combines the strengths of the Swin transformer, genetic algorithm (GA) feature selection, and the replacement of the softmax layer with a random forest classifier. Our methodology involves extracting feature representations from the Swin transformer, utilizing GA to identify the optimal feature set, and employing random forest as the classification model. Additionally, data augmentation techniques are applied to augment the diversity and quantity of the SIPaKMeD1 cervical cancer image dataset. We compare the performance of the Swin-GA-RF Transformer with pre-trained CNN models using two classes and five classes of cervical cancer classification, employing both Adam and SGD optimizers. The experimental results demonstrate that Swin-GA-RF outperforms other Swin transformers and pre-trained CNN models. When utilizing the Adam optimizer, Swin-GA-RF achieves the highest performance in both binary and five-class classification tasks. Specifically, for binary classification, it achieves an accuracy, precision, recall, and F1-score of 99.012, 99.015, 99.012, and 99.011, respectively. In the five-class classification, it achieves an accuracy, precision, recall, and F1-score of 98.808, 98.812, 98.808, and 98.808, respectively. These results underscore the effectiveness of the Swin-GA-RF approach in cervical cancer classification, demonstrating its potential as a valuable tool for early diagnosis and screening programs.

CloudMix: Dual Mixup Consistency for Unpaired Point Cloud Completion.

Due to the unsatisfactory performance of supervised methods on unpaired real-world scans, point cloud completion via cross-domain adaptation has recently drawn growing attention. Nevertheless, previous approaches only focus on alleviating the distribution shift through domain alignment, resulting in massive information loss of real-world domain data. To tackle this issue, we propose a dual mixup-induced consistency regularization to integrate both source and target domain to improve robustness and generalization capability. Specifically, we mix up virtual and real-world shapes in the input and latent feature space respectively, and then regularize the completion network by forcing two kinds of mixed completion predictions to be consistent. To further adapt to each instance within the real-world domain, we design a novel density-aware refiner to utilize local context information to preserve the fine-grained details and remove noise or outliers for coarse completion. Extensive experiments on real-world scans and our synthetic unpaired datasets demonstrate the superiority of our method over existing state-of-the-art approaches.

COSTA: A Multi-center TOF-MRA Dataset and A Style Self-Consistency Network for Cerebrovascular Segmentation.

Time-of-flight magnetic resonance angiography (TOF-MRA) is the least invasive and ionizing radiation-free approach for cerebrovascular imaging, but variations in imaging artifacts across different clinical centers and imaging vendors result in inter-site and inter-vendor heterogeneity, making its accurate and robust cerebrovascular segmentation challenging. Moreover, the limited availability and quality of annotated data pose further challenges for segmentation methods to generalize well to unseen datasets. In this paper, we construct the largest and most diverse TOF-MRA dataset (COSTA) from 8 individual imaging centers, with all the volumes manually annotated. Then we propose a novel network for cerebrovascular segmentation, namely CESAR, with the ability to tackle feature granularity and image style heterogeneity issues. Specifically, a coarse-to-fine architecture is implemented to refine cerebrovascular segmentation in an iterative manner. An automatic feature selection module is proposed to selectively fuse global long-range dependencies and local contextual information of cerebrovascular structures. A style self-consistency loss is then introduced to explicitly align diverse styles of TOF-MRA images to a standardized one. Extensive experimental results on the COSTA dataset demonstrate the effectiveness of our CESAR network against state-of-the-art methods. We have made 6 subsets of COSTA with the source code online available, in order to promote relevant research in the community.

Knowledge distillation in transformers with tripartite attention: Multiclass brain tumor detection in highly augmented MRIs

The advent of attention-based architectures in medical imaging has ushered in an era of precision diagnostics, particularly in the detection and classification of brain tumors. This study introduced an innovative knowledge distillation framework employing a tripartite attention mechanism within transformer encoder models, specifically tailored for the identification of multiple brain tumor classes through magnetic resonance imaging (MRI). The proposed methodology synergistically harnesses the capabilities of large, highly parameterized teacher models to train more compact, efficient student models suitable for deployment in resource-constrained environments such as the internet of medical things and smart healthcare devices. Utilizing a diverse array of MRI sequences—including T1, contrast-enhanced T1, and T2—this study accounts for the nuanced variations across brain tumor classes derived from three extensive datasets. The tripartite attention mechanism addresses the limitation of traditional attention models by innovatively integrating temperature-softening neighborhood attention, global attention, and cross-attention layers. This sophisticated approach allows for a richer and more nuanced feature representation, capturing both local and global contextual information and intricate tumor features within MRI scans. This is supplemented by a unique augmentation pipeline and shifted patch tokenization technique, which enrich the model's input representation, especially for underrepresented classes. Through meticulous experimentation and ablation studies, the study demonstrates that the proposed model not only retains the robustness of its larger counterparts but also delivers enhanced performance metrics. When juxtaposed with benchmarking models—including traditional deep CNNs and various transformer-based architectures—the proposed model consistently showcases superior results. Its effectiveness is reflected in its lower teacher and student losses, commendable Brier scores, and noteworthy top-1 and top-5 accuracies, as well as AUC metrics across all datasets. This paper not only validates the efficacy of knowledge distillation in complex medical image analysis tasks but also provides a promising pathway for the integration of cutting-edge AI techniques in real-world clinical applications, potentially revolutionizing the early detection and treatment of brain tumors.

NnUnetFormer: an automatic method based on nnUnet and transformer for brain tumor segmentation with multimodal MR images

Objective. Both local and global context information is crucial semantic features for brain tumor segmentation, while almost all the CNN-based methods cannot learn global spatial dependencies very well due to the limitation of convolution operations. The purpose of this paper is to build a new framework to make full use of local and global features from multimodal MR images for improving the performance of brain tumor segmentation. Approach. A new automated segmentation method named nnUnetFormer was proposed based on nnUnet and transformer. It fused transformer modules into the deeper layers of the nnUnet framework to efficiently obtain both local and global features of lesion regions from multimodal MR images. Main results. We evaluated our method on BraTS 2021 dataset by 5-fold cross-validation and achieved excellent performance with Dice similarity coefficient (DSC) 0.936, 0.921 and 0.872, and 95th percentile of Hausdorff distance (HD95) 3.96, 4.57 and 10.45 for the regions of whole tumor (WT), tumor core (TC), and enhancing tumor (ET), respectively, which outperformed recent state-of-the-art methods in terms of both average DSC and average HD95. Besides, ablation experiments showed that fusing transformer into our modified nnUnet framework improves the performance of brain tumor segmentation, especially for the TC region. Moreover, for validating the generalization capacity of our method, we further conducted experiments on FeTS 2021 dataset and achieved satisfactory segmentation performance on 11 unseen institutions with DSC 0.912, 0.872 and 0.759, and HD95 6.16, 8.81 and 38.50 for the regions of WT, TC, and ET, respectively. Significance. Extensive qualitative and quantitative experimental results demonstrated that the proposed method has competitive performance against the state-of-the-art methods, indicating its interest for clinical applications.

Dynamic context modeling based lightweight high-resolution network for dense prediction

Recent research shows that high-resolution networks can provide representative multi-scale features for vision tasks. However, high-resolution-based architectures are weak in capturing spatial long-range information, and they are computationally expensive. To alleviate these problems and encourage the engineering applications of dense prediction, based on vanilla high-resolution network, this paper presents a novel efficient architecture for dense prediction, namely Dynamic context modeling based lightweight High-Resolution Network (Dynamic-HRNet). In particular, the network consists of two key components: (i) dynamic split convolution, a more efficient and flexible convolution compared to the standard convolution, which could be conveniently embedded into other networks, and (ii) adaptive context modeling, which adaptively captures local and global contextual information to enrich the representation in parallel. Using the above two components, two lightweight blocks are designed that serve as the basic building units of Dynamic-HRNet, termed dynamic multi-scale context block and dynamic global context block, respectively. Extensive experiments demonstrate that the proposed Dynamic-HRNet, as a backbone, achieves significant superiority in popular benchmarks for human pose estimation (70.6% AP on COCO dataset and 87.6% PCKh on MPII dataset) and semantic segmentation (75.3% MIoU on Cityscapes dataset). Considering both efficiency and accuracy, it outperforms the most advanced lightweight architectures.

Global semantic-guided network for saliency prediction

The human visual system effectively analyzes scenes based on local, global and semantic properties. Deep learning-based saliency prediction models adopted two-stream networks, leveraged prior knowledge of global semantics, or added long-range dependency modeling structures like transformers to incorporate global saliency information. However, they either brought high complexity to learning local and global features or neglected the design for enhancing local features. In this paper, we propose a Global Semantic-Guided Network (GSGNet), which first enriches global semantics through a modified transformer block and then incorporates semantic information into visual features from local and global perspectives in an efficient way. Multi-head self-attention in transformers captures global features, but lacks information communication within and between feature subspaces (heads) when computing the similarity matrix. To learn global representations and enhance interactions of the subspaces, we propose a Channel-Squeeze Spatial Attention (CSSA) module to emphasize channel-relevant information in a compression manner and learn global spatial relationships. To better fuse local and global contextual information, we propose a hybrid CNN-Transformer block called local–global fusion block (LGFB) for aggregating semantic features simply and efficiently. Experimental results on four public datasets demonstrate that our model achieves compelling performance compared with the state-of-the-art saliency prediction models on various evaluation metrics.

Weed–Crop Segmentation in Drone Images with a Novel Encoder–Decoder Framework Enhanced via Attention Modules

The rapid expansion of the world’s population has resulted in an increased demand for agricultural products which necessitates the need to improve crop yields. To enhance crop yields, it is imperative to control weeds. Traditionally, weed control predominantly relied on the use of herbicides; however, the indiscriminate application of herbicides presents potential hazards to both crop health and productivity. Fortunately, the advent of cutting-edge technologies such as unmanned vehicle technology (UAVs) and computer vision has provided automated and efficient solutions for weed control. These approaches leverage drone images to detect and identify weeds with a certain level of accuracy. Nevertheless, the identification of weeds in drone images poses significant challenges attributed to factors like occlusion, variations in color and texture, and disparities in scale. The utilization of traditional image processing techniques and deep learning approaches, which are commonly employed in existing methods, presents difficulties in extracting features and addressing scale variations. In order to address these challenges, an innovative deep learning framework is introduced which is designed to classify every pixel in a drone image into categories such as weed, crop, and others. In general, our proposed network adopts an encoder–decoder structure. The encoder component of the network effectively combines the Dense-inception network with the Atrous spatial pyramid pooling module, enabling the extraction of multi-scale features and capturing local and global contextual information seamlessly. The decoder component of the network incorporates deconvolution layers and attention units, namely, channel and spatial attention units (CnSAUs), which contribute to the restoration of spatial information and enhance the precise localization of weeds and crops in the images. The performance of the proposed framework is assessed using a publicly available benchmark dataset known for its complexity. The effectiveness of the proposed framework is demonstrated via comprehensive experiments, showcasing its superiority by achieving a 0.81 mean Intersection over Union (mIoU) on the challenging dataset.

Deep understanding of radiology reports: leveraging dynamic convolution in chest X-ray images

PurposeThe purpose of this study is to investigate and demonstrate the advancements achieved in the field of chest X-ray image captioning through the utilization of dynamic convolutional encoder–decoder networks (DyCNN). Typical convolutional neural networks (CNNs) are unable to capture both local and global contextual information effectively and apply a uniform operation to all pixels in an image. To address this, we propose an innovative approach that integrates a dynamic convolution operation at the encoder stage, improving image encoding quality and disease detection. In addition, a decoder based on the gated recurrent unit (GRU) is used for language modeling, and an attention network is incorporated to enhance consistency. This novel combination allows for improved feature extraction, mimicking the expertise of radiologists by selectively focusing on important areas and producing coherent captions with valuable clinical information.Design/methodology/approachIn this study, we have presented a new report generation approach that utilizes dynamic convolution applied Resnet-101 (DyCNN) as an encoder (Verelst and Tuytelaars, 2019) and GRU as a decoder (Dey and Salemt, 2017; Pan et al., 2020), along with an attention network (see Figure 1). This integration innovatively extends the capabilities of image encoding and sequential caption generation, representing a shift from conventional CNN architectures. With its ability to dynamically adapt receptive fields, the DyCNN excels at capturing features of varying scales within the CXR images. This dynamic adaptability significantly enhances the granularity of feature extraction, enabling precise representation of localized abnormalities and structural intricacies. By incorporating this flexibility into the encoding process, our model can distil meaningful and contextually rich features from the radiographic data. While the attention mechanism enables the model to selectively focus on different regions of the image during caption generation. The attention mechanism enhances the report generation process by allowing the model to assign different importance weights to different regions of the image, mimicking human perception. In parallel, the GRU-based decoder adds a critical dimension to the process by ensuring a smooth, sequential generation of captions.FindingsThe findings of this study highlight the significant advancements achieved in chest X-ray image captioning through the utilization of dynamic convolutional encoder–decoder networks (DyCNN). Experiments conducted using the IU-Chest X-ray datasets showed that the proposed model outperformed other state-of-the-art approaches. The model achieved notable scores, including a BLEU_1 score of 0.591, a BLEU_2 score of 0.347, a BLEU_3 score of 0.277 and a BLEU_4 score of 0.155. These results highlight the efficiency and efficacy of the model in producing precise radiology reports, enhancing image interpretation and clinical decision-making.Originality/valueThis work is the first of its kind, which employs DyCNN as an encoder to extract features from CXR images. In addition, GRU as the decoder for language modeling was utilized and the attention mechanisms into the model architecture were incorporated.

Point cloud semantic segmentation based on local feature fusion and multilayer attention network

AbstractSemantic segmentation from a three‐dimensional point cloud is vital in autonomous driving, computer vision, and augmented reality. However, current semantic segmentation does not effectively use the point cloud's local geometric features and contextual information, essential for improving segmentation accuracy. A semantic segmentation network that uses local feature fusion and a multilayer attention mechanism is proposed to address these challenges. Specifically, the authors designed a local feature fusion module to encode the geometric and feature information separately, which fully leverages the point cloud's feature perception and geometric structure representation. Furthermore, the authors designed a multilayer attention pooling module consisting of local attention pooling and cascade attention pooling to extract contextual information. Local attention pooling is used to learn local neighbourhood information, and cascade attention pooling captures contextual information from deeper local neighbourhoods. Finally, an enhanced feature representation of important information is obtained by aggregating the features from the two deep attention pooling methods. Extensive experiments on large‐scale point‐cloud datasets Stanford 3D large‐scale indoor spaces and SemanticKITTI indicate that authors network shows excellent advantages over existing representative methods regarding local geometric feature description and global contextual relationships.

CFA-GAN: Cross fusion attention and frequency loss for image style transfer

Image style transfer is a significant challenge in computer vision, where the goal is to transfer the style of a reference image to a source image while preserving its background and feature information. Despite previous progress in this area, many studies have overlooked the issues of local details and complexity, leading to unrealistic generated images and slow model inference speeds. This paper proposes an efficient image style transfer method based on cross-fusion attention (CFA), inspired by generative adversarial networks (GAN) and a new joint loss. Our novel CFA encodes local contextual information between low-level and high-level features, models remote dependencies of features, and embeds CFA into the encoder of our proposed model to increase the diversity of small features and focus on more prominent feature parts. Furthermore, we use frequency domain loss to improve image reconstruction quality and a joint loss to train the network. Our experiments on multiple datasets demonstrate that our proposed method offers significant advantages in terms of inference speed, background information preservation, and local details, making it a promising approach to image style transfer. The code for our method is available at https://github.com/berylxzhang/CFA-GAN.

A dual-stream recurrence-attention network with global–local awareness for emotion recognition in textual dialog

In real-world dialog systems, the ability to understand the user’s emotions and interact anthropomorphically is of great significance. Emotion Recognition in Conversation (ERC) is one of the key ways to accomplish this goal and has attracted growing attention. How to model the context in a conversation is a central aspect and a major challenge of ERC tasks. Most existing approaches struggle to adequately incorporate both global and local contextual information, and their network structures are overly sophisticated. For this reason, we propose a simple and effective Dual-stream Recurrence-Attention Network (DualRAN), which is based on Recurrent Neural Network (RNN) and Multi-head ATtention network (MAT). DualRAN eschews the complex components of current methods and focuses on combining recurrence-based methods with attention-based ones. DualRAN is a dual-stream structure mainly consisting of local- and global-aware modules, modeling a conversation simultaneously from distinct perspectives. In addition, we develop two single-stream network variants for DualRAN, i.e., SingleRANv1 and SingleRANv2. According to the experimental findings, DualRAN boosts the weighted F1 scores by 1.43% and 0.64% on the IEMOCAP and MELD datasets, respectively, in comparison to the strongest baseline. On two other datasets (i.e., EmoryNLP and DailyDialog), our method also attains competitive results.

Memory-efficient transformer network with feature fusion for breast tumor segmentation and classification task

The analysis of breast cancer using Ultrasounds, Magnetic resonance imaging (MRI), and Mammogram images plays a crucial role in the early detection of breast tumors in women. Nevertheless, accurately segmenting breast tumors remains a challenging task due to the presence of random variations, irregular shapes, and blurred boundaries within tumor regions. Blurred boundaries between breast tumor regions and healthy tissues may also cause higher rates of false-positives or incorrect segmentation results. Traditional convolutional neural networks (CNNs) are constrained by their inability to effectively capture global context information. As a result, they often yield unsatisfactory results in many scenarios. To address this, we have proposed a double encoder block to capture local and global context information related to breast tumor segmentation. In this model, we used a memory-efficient decoder block and a feature fusion block (FFB) to improve feature adaptation and classification capability among adjacent-level with the help of activating channels. Pyramid-dilated fusion (PDF) block based on depth-wise separable convolution is integrated within the transformer to achieve a comprehensive multiscale context. We enabled our model to suppress the irrelevant background noises of breast ultrasound and MRI images. Following this, we conducted comprehensive experiments to evaluate the effectiveness of our proposed Memory-efficient transformer network (MET-Net) for breast tumor segmentation and classification. To ensure reliable results, we employed ultrasounds and MRI datasets within a simulation environment. During the experiment analysis, the MET-Net model demonstrated extraordinary results in the presence of existing state-of-the-art CNNs models in terms of comparative measures. Source code of this work is available at https://github.com/ahmedeqbal/MET-Net.

SignalFormer: Hybrid Transformer for Automatic Drone Identification Based on Drone RF Signals.

With the growing integration of drones into various civilian applications, the demand for effective automatic drone identification (ADI) technology has become essential to monitor malicious drone flights and mitigate potential threats. While numerous convolutional neural network (CNN)-based methods have been proposed for ADI tasks, the inherent local connectivity of the convolution operator in CNN models severely constrains RF signal identification performance. In this paper, we propose an innovative hybrid transformer model featuring a CNN-based tokenization method that is capable of generating T-F tokens enriched with significant local context information, and complemented by an efficient gated self-attention mechanism to capture global time/frequency correlations among these T-F tokens. Furthermore, we underscore the substantial impact of incorporating phase information into the input of the SignalFormer model. We evaluated the proposed method on two public datasets under Gaussian white noise and co-frequency signal interference conditions, The SignalFormer model achieved impressive identification accuracy of 97.57% and 98.03% for coarse-grained identification tasks, and 97.48% and 98.16% for fine-grained identification tasks. Furthermore, we introduced a class-incremental learning evaluation to demonstrate SignalFormer's competence in handling previously unseen categories of drone signals. The above results collectively demonstrate that the proposed method is a promising solution for supporting the ADI task in reliable ways.