The effectiveness of the SAR object detection technique based on Convolutional Neural Networks (CNNs) has been widely proven, and it is increasingly used in the recognition of ship targets. Recently, efforts have been made to integrate transformer structures into SAR detectors to achieve improved target localization. However, existing methods rarely design the transformer itself as a detector, failing to fully leverage the long-range modeling advantages of self-attention. Furthermore, there has been limited research into multi-class SAR target detection. To address these limitations, this study proposes a SAR detector named CCDN-DETR, which builds upon the framework of the detection transformer (DETR). To adapt to the multiscale characteristics of SAR data, cross-scale encoders were introduced to facilitate comprehensive information modeling and fusion across different scales. Simultaneously, we optimized the query selection scheme for the input decoder layers, employing IOU loss to assist in initializing object queries more effectively. Additionally, we introduced constrained contrastive denoising training at the decoder layers to enhance the model’s convergence speed and improve the detection of different categories of SAR targets. In the benchmark evaluation on a joint dataset composed of SSDD, HRSID, and SAR-AIRcraft datasets, CCDN-DETR achieves a mean Average Precision (mAP) of 91.9%. Furthermore, it demonstrates significant competitiveness with 83.7% mAP on the multi-class MSAR dataset compared to CNN-based models.

Recurrent Neural Networks (RNNs) are classical models for processing sequential data, demonstrating excellent performance in tasks such as natural language processing and time series prediction. However, during the training of RNNs, the issues of vanishing and exploding gradients often arise, significantly impacting the model’s performance and efficiency. In this paper, we investigate why RNNs are more prone to gradient problems compared to other common sequential networks. To address this issue and enhance network performance, we propose a method for gradient normalization of network weights. This method suppresses the occurrence of gradient problems by altering the statistical properties of RNN weights, thereby improving training effectiveness. Additionally, we analyze the impact of weight gradient normalization on the probability-distribution characteristics of model weights and validate the sensitivity of this method to hyperparameters such as learning rate. The experimental results demonstrate that gradient normalization enhances the stability of model training and reduces the frequency of gradient issues. On the Penn Treebank dataset, this method achieves a perplexity level of 110.89, representing an 11.48% improvement over conventional gradient descent methods. For prediction lengths of 24 and 96 on the ETTm1 dataset, Mean Absolute Error (MAE) values of 0.778 and 0.592 are attained, respectively, resulting in 3.00% and 6.77% improvement over conventional gradient descent methods. Moreover, selected subsets of the UCR dataset show an increase in accuracy ranging from 0.4% to 6.0%. The gradient normalization method enhances the ability of RNNs to learn from sequential and causal data, thereby holding significant implications for optimizing the training effectiveness of RNN-based models.

Resource-constrained Internet of Things (IoT) devices demand efficient and robust intrusion detection systems (IDSs) to counter evolving cyber threats. The traditional IDS models, however, struggle with high computational complexity and inadequate feature extraction, limiting their accuracy and generalizability in IoT environments. To address this, we propose FFT-RDNet, a lightweight IDS framework leveraging depthwise separable convolution and frequency-domain feature fusion. An ADASYN-Tomek Links hybrid strategy first addresses class imbalances. The core innovation of FFT-RDNet lies in its novel two-dimensional spatial feature modeling approach, realized through a dedicated dual-path feature embedding module. One branch extracts discriminative statistical features in the time domain, while the other branch transforms the data into the frequency domain via Fast Fourier Transform (FFT) to capture the essential energy distribution characteristics. These time–frequency domain features are fused to construct a two-dimensional feature space, which is then processed by a streamlined residual network using depthwise separable convolution. This network effectively captures complex periodic attack patterns with minimal computational overhead. Comprehensive evaluation on the NSL-KDD and CIC-IDS2018 datasets shows that FFT-RDNet outperforms state-of-the-art neural network IDSs across accuracy, precision, recall, and F1 score (improvements: 0.22–1%). Crucially, it achieves superior accuracy with a significantly reduced computational complexity, demonstrating high efficiency for resource-constrained IoT security deployments.

With the rapid increase in cyber-attacks, intrusion detection systems (IDS) have become essential for network security. However, traditional IDS methods often struggle with class imbalance, leading to asymmetric data distributions that adversely affect detection performance and model generalization. To address this issue and enhance detection accuracy, this paper proposes SE-DWNet, a residual network model incorporating an attention mechanism and one-dimensional depthwise separable convolution, trained on a symmetrically preprocessed dataset using SMOTETomek sampling. First, the feature distributions of the training and test datasets are analyzed using box plots, highlighting the impact of feature difference. To mitigate this difference and restore a more symmetric data distribution, we employ the SMOTETomek integrated sampling method in conjunction with a Focal Loss function. Subsequently, a lightweight residual network, incorporating the SE module and the Res-DWNet module, is designed to improve detection accuracy while maintaining computational efficiency. Extensive experiments on the NSL-KDD, CICIDS2018, and ToN-IoT datasets demonstrate that SE-DWNet outperforms existing neural network-based IDS models, achieving accuracy, precision, recall, and F1-score improvements ranging from 0.17% to 5.33%. The results confirm the effectiveness and superiority of the proposed approach in intrusion detection tasks.

Common human pose estimation methods rely on 2D heatmap regression, which requires expensive upsampling layers to maintain the resolution of the heatmap and additional post-processing for coordinate decoding. These components hinder the inference speed of human pose estimation tasks. To address this challenge, we propose a new real-time human pose estimation framework, EdgePose. First, we design the convolutional module EdgeBlock-C and the edge attention module EdgeBlock-T, and then build a hybrid network based on them to take advantage of both ConvNet and VIT. In addition, EdgePose simplifies the human pose estimation process by converting the output of the key point coordinates into a pixel classification task along the horizontal and vertical axes, thereby eliminating the upsampling and post-processing operations that may hinder inference speed, and speeding up the model’s inference speed while ensuring accuracy. The experimental results show that EdgePose-S achieved an AP score of 68.6 in the COCO validation test, and at the same time achieved an inference speed of 285.7 FPS on an Intel i9-10920X CPU. In the embedded Jetson Xavier NX environment, EdgePose-B achieved an AP score of 72.2 and an inference speed of 51.3 FPS, which is better than existing two-stage pose estimation methods.