Due to the limited annotated data, semantic segmentation faces significant challenges in few shot generalization. Current methods combining prototype learning and affinity learning suffer from static prototypes failing to capture intra-class diversity. Even if these issues are addressed, they still encounter the critical problem of ineffective integration between global information acquired through prototype learning and local information obtained via affinity learning, which ultimately hinders their performance in few-shot segmentation scenarios. Inspired by human cognitive mechanisms, we propose CCANet - a two-stage framework that progressively combines prototype learning with affinity learning through "category-agnostic to category-aware" knowledge evolution. This innovative architecture effectively resolves the aforementioned limitations while achieving systematic knowledge integration. The Global category-aware Module employs category-agnostic localization and dynamic prototype evolution, leveraging category-agnostic global context modeling and history-driven prototype adaptation to enhance cross-category alignment and noise robustness. The Local Feature Enrichment Module introduces a self-attention-gated fusion paradigm, coordinating multi-scale global-local correlations to effectively guide the affinity modeling process across multi-scale contexts. Extensive experiments demonstrate state-of-the-art performance on PASCAL-5i and COCO-20i, validating the effectiveness of our approach. Our work bridges cognitive principles with computational models, offering new insights for few-shot segmentation.

ABSTRACT With the advancement of deep learning, single‐image super‐resolution (SISR) has achieved significant progress. Recently, vision transformer‐based super‐resolution models have demonstrated remarkable performance; however, their high computational cost hinders their practical application. In this paper, we introduce a lightweight transformer‐based super‐resolution model termed information‐enhanced efficient multi‐directional transformer(IEMT). The model employs a dual‐branch architecture that integrates the strengths of both convolutional neural network (CNN) and transformer networks. The proposed high‐frequency extraction block (HEB) effectively captures high‐frequency information from the enhanced image. Furthermore, a multi‐directional attention mechanism is incorporated into the transformer branch to comprehensively learn latent features and details, thereby enhancing reconstruction quality. For attention computation, we propose a dynamic parameter‐sharing mechanism that adaptively adjusts parameter sharing based on local image features, significantly reducing the model's parameter count. Experimental results demonstrate that the proposed IEMT achieves superior performance on five benchmark datasets, with a significantly reduced parameter count, computational complexity, and memory usage.

Abstract Existing pre-trained models have yielded promising results in terms of computational time reduction. However, these models only focus on pruning simple sentences or less salient words, while neglecting the treatment of relatively complex sentences. It is frequently these sentences that cause the loss of model accuracy. This shows that the adaptation of the existing models is one-sided. To address this issue, in this paper, we propose a sample-adaptive training and inference model. Specifically, complex samples are extracted from the training datasets and a dedicated data augmentation module is trained to extract global and local semantic information of complex samples. During inference, simple samples can exit the model via the Sample Adaptive Exit Mechanism, Normal samples pass through the whole backbone model before inference, while complex samples are processed by the Characteristic Enhancement Module after passing through the backbone model. In this way, all samples are processed adaptively. Our extensive experiments on classification tasks datasets in the field of Natural Language Processing demonstrate that our method enhances model accuracy and reduces model inference time for multiple datasets. Moreover, our method is transferable and can be applied to multiple pre-trained models.

Abstract Few‐shot segmentation (FSS) enables rapid adaptation to the segmentation task of unseen‐classes object based on a few labelled support samples. Currently, the focal point of research in the FSS field is to align features between support and query images, aiming to improve the segmentation performance. However, most existing FSS methods implement such support/query alignment by solely leveraging middle‐level feature for generalization, ignoring the category semantic information contained in high‐level feature, while pooling operation inevitably lose spatial information of the feature. To alleviate these issues, the authors propose the Iterative Segmentation Network Based on Feature Pyramid (FPIseg), which mainly consists of three modules: Feature Pyramid Fusion Module (FPFM), Region Feature Enhancement Module (RFEM), and Iterative Optimization Segmentation Module (IOSM). Firstly, FPFM fully utilizes the foreground information from the support image to implement support/query alignment under multi‐scale, multi‐level semantic backgrounds. Secondly, RFEM enhances the foreground detail information of aligned feature to improve generalization ability. Finally, ISOM iteratively segments the query image to optimize the prediction result and improve segmentation performance. Extensive experiments on the PASCAL‐5 i and COCO‐20 i datasets show that FPIseg achieves considerable segmentation performance under both 1‐shot and 5‐shot settings.

Abstract Recently, convolutional neural network (CNN) has shown great power in single image super resolution (SISR) reconstruction, and achieving significant improvements over traditional methods. Despite the great success of these CNN‐based methods, direct application of these methods to some edge devices is impractical due to the large computational overhead required. To address this problem, a novel, lightweight SISR network focusing on speed and accuracy, called the multi‐path feedback fusion network (MFFN), has been designed in this paper. Specifically, in order to extract features more effectively, the authors propose a novel fusion attention feedback block (FAFB) as the main building block of MFFN. The FAFB consists of a backbone branch and several hierarchical branches. The backbone branch is composed of stacked enhanced pixel attentional blocks (EPAB), which are responsible for incremental deep feature learning on the feature map. And the hierarchical branches are responsible for extracting feature maps with different sizes of receptive fields and fusing these feature maps with the features extracted from the trunk branches to achieve multi‐scale feature learning, which the authors refer to this design as the multi‐scale fusion block (MSFB). Extra enhancement information (EIE) is added to each EPAB input, which enables the backbone branch to learn more effectively. On the other hand, the outputs of the cascade branches are further complemented by an additional feedback fusion enhancement block (FFEB) before being fused with the output of the trunk branches to achieve more comprehensive and accurate feature learning. Numerous experiments have shown that MFFN achieves higher accuracy than other state‐of‐the‐art methods on benchmark test sets.

This paper proposes a new prediction-based dynamic multiobjective optimization (PBDMO) method, which combines a new prediction-based reaction mechanism and a popular regularity model-based multiobjective estimation of distribution algorithm (RM-MEDA) for solving dynamic multiobjective optimization problems. Whenever a change is detected, PBDMO reacts effectively to it by generating three subpopulations based on different strategies. The first subpopulation is created by moving nondominated individuals using a simple linear prediction model with different step sizes. The second subpopulation consists of some individuals generated by a novel sampling strategy to improve population convergence as well as distribution. The third subpopulation comprises some individuals generated using a shrinking strategy based on the probability distribution of variables. These subpopulations are tailored to form a population for the new environment. The experimental results carried out on a variety of bi- and three-objective benchmark functions demonstrate that the proposed technique has competitive performance compared with some state-of-the-art algorithms.