Few-Shot Semantic Segmentation (FSS) effectively segments new classes with limited data. However, the often-overlooked style differences between support and query sets can lead to feature shifts, disrupting accurate feature matching due to inadequate abstraction in mid-level features. To tackle this challenge, we introduce a novel network for disentangling style representations from a frequency perspective. Specifically, we introduce a parameter-free Adaptive Style Fourier Alignment module that performs regional frequency replacement to generate style-aligned pseudo-support images. To further refine style adaptation, we construct Style-Aware Prototypes and employ a Style Modulation Module that selectively adjusts the query features based on low-frequency modulation via wavelet transform to preserve edge details. Extensive experiments on benchmark datasets demonstrate the effectiveness of our approach, yielding mIoU improvements of 1.91% in the 1-shot setting and 1.54% in the 5-shot configuration.

Unsupervised Domain Adaptation (UDA) leverages labeled data from the source domain to align the target domain distribution, yet its performance is constrained by the lack of supervision in the target domain, resulting in a significant performance gap compared to fully supervised methods. Semi-Supervised Learning (SSL) combines limited labeled data with abundant unlabeled data but assumes identical distributions between labeled and unlabeled data, failing to address cross-domain challenges in medical imaging. To address these limitations, this paper proposes a Semi-Supervised Domain Adaptation (SSDA) framework based on Local-Global Hybridization and Dual-Teacher Collaborative Distillation, aiming to enhance the robustness and accuracy of medical image segmentation. The main contributions are summarized as follows:(1) FMix is introduced to utilize frequency-domain information for generating mask regions simulating real organ structures, enabling the model to focus on local features;(2) Mixup is incorporated to enhance domain invariance by simulating inter-domain transitions, guiding the model to prioritize global features;(3) A complementary dual-teacher distillation model is constructed, leveraging FMix-Mixup co-enhanced training to synergize local-global feature learning;(4) Monte Carlo Dropout is adopted to filter low-confidence pseudo-labels and iteratively update the target-domain labeled dataset. Extensive experiments on the BraTS2018 benchmark dataset demonstrate that the proposed framework significantly improves cross-modality medical image segmentation performance. With only one labeled target sample, it achieves over 10% improvement in Dice coefficient. Compared to state-of-the-art SSDA methods, the framework achieves a 4.7% higher Dice score, approaching the performance of fully supervised learning.

This paper introduces a lightweight spectral channel feature transformation network, LDCFormer, designed to address the high computational complexity and excessive parameter count resulting from self-attention and spatial MLP (Multilayer Perceptron) in vision transformers when dealing with long sequences. Initially, the image information is transformed into the frequency domain using a two-dimensional discrete cosine transform (DCT), effectively capturing the image’s frequency domain features. Secondly, considering that different frequency areas represent various types of features, local feature information such as edges and textures are extracted in the high-frequency area, while the image’s global feature information is extracted in the low-frequency area, complemented by channel attention for feature cleansing. Finally, the integration and interaction of global feature information in the image are achieved by introducing the Transformer architecture. LDCFormer employs a zero-learning-parameter 2D LDCFormer operation to extract features directly from the frequency domain, significantly reducing the number of trainable parameters, and utilizes depth-separable LDConv MLP to further accelerate computational speed, achieving the lightweight and efficient characteristics of LDCFormer. Accuracies of 78.8%, 88.9% and 88.6% were attained on three typical datasets. The experimental results demonstrate that LDCFormer maintains high classification performance while reducing the parameter count, achieving a good balance between speed and accuracy.

The fog computing testbed can accelerate the development of fog computing. But so far, there is no real dedicated fog computing testbed to help researchers to test their prototypes, designs, and distributed algorithms in real fog computing scenarios. Researchers often verify their fog computing solutions by adapting some existing testbeds or using simulators under specific conditions, which may make their experimental results unable to withstand the examination of real production environments. This paper proposes piFogBed, the first fog computing testbed for rapid prototyping fog computing components in real environments by using Raspberry Pies and the Docker container. It can be integrated into the existing production network quickly and flexibly. To evaluate piFogBed, we demonstrate its experimental process with a medical monitoring use case, and also verify its feasibility, effectiveness and fidelity. The cost of each fog node of the proposed testbed is only 400 RMB, which is very low compared with the special fog computing equipment, so it is worth promoting and applying.

Various networking applications and systems must be tested before the final deployment. Many of the tests are performed on network testbeds such as Emulab, PlanetLab, etc. These testbeds are large in scale and organize devices in relatively fixed ways. It is difficult for them to incorporate the latest personalized devices, such as smart watches, smart glasses and other emerging gadgets, so they tend to fall short in supporting personalized experiments using devices around users. Moreover, these testbeds commonly impose restrictions on users in terms of when and where to carry out experiments making them clumsy or inconvenient to use. This paper proposes to build a testbed using users' devices in their own personal area networks (PANs). We have designed and implemented a prototype of this system, which we call PANBED. Our experiments show that PANBED allows users to set up different scenes to test applications using a home router, PCs, mobile phones and other equipment. PANBED is light weighted with a size less than 16 KB and it has little impact to the other functions of the PAN. When one node keeps on sending 32-bytes packets to another for 30 seconds, PANBED exhibited little impact on the memory of the router, and the CPU load of the router was always less than 25%.