Multimodal sentiment analysis has gained popularity as a research field for its ability to predict users’ emotional tendencies more comprehensively. The data fusion module is a critical component of multimodal sentiment analysis, as it allows for integrating information from multiple modalities. However, it is challenging to combine modalities and remove redundant information effectively. In our research, we address these challenges by proposing a multimodal sentiment analysis model based on supervised contrastive learning, which leads to more effective data representation and richer multimodal features. Specifically, we introduce the MLFC module, which utilizes a convolutional neural network (CNN) and Transformer to solve the redundancy problem of each modal feature and reduce irrelevant information. Moreover, our model employs supervised contrastive learning to enhance its ability to learn standard sentiment features from data. We evaluate our model on three widely-used datasets, namely MVSA-single, MVSA-multiple, and HFM, demonstrating that our model outperforms the state-of-the-art model. Finally, we conduct ablation experiments to validate the efficacy of our proposed method.

Infrared small target detection (ISTD) has a wide range of applications in both military and civilian fields. Due to their low contrast and the absence of color and texture information, small targets in infrared images can be readily obscured by complex background clutter. Most existing methods primarily concentrate on local modeling, but they lack sufficient feature interactions, leading to the loss of valuable feature information. Hence, we propose a straightforward and efficient cross-layer feature guided multiscale network (CMNet) for the ISTD. We devise the adjacent layer feature guidance strategy (ALFG) to boost the expressiveness of low-level features through the utilization of semantic guidance from high-level features, enabling a more effective capture of target feature, shape, and structure information. Furthermore, we introduce the multiscale residual connection block (MRCB) that thoroughly leverages multiscale feature information, consequently augmenting perceptual ability and feature representation. For our proposed CMNet, the Intersection over Union (IoU) on the NUAA-SIRST and NUDT-SIRST datasets is 83.32% and 95.74%, while the normalized Intersection over Union (nIoU) is 83.37% and 95.61%, respectively. CMNet exhibits markedly superior detection performance compared to other networks for infrared small target detection. The code is available at github.com/YuanMortal/CMNet.

Camouflaged object detection (COD) aims to segment those camouflaged objects that blend perfectly into their surroundings. Due to the low boundary contrast between camouflaged objects and their surroundings, their detection poses a significant challenge. Despite the numerous excellent camouflaged object detection methods developed in recent years, issues such as boundary refinement and multi-level feature extraction and fusion still need further exploration. In this paper, we propose a novel multi-level feature integration network (MFNet) for camouflaged object detection. Firstly, we design an edge guidance module (EGM) to improve the COD performance by providing additional boundary semantic information by combining high-level semantic information and low-level spatial details to model the edges of camouflaged objects. Additionally, we propose a multi-level feature integration module (MFIM), which leverages the fine local information of low-level features and the rich global information of high-level features in adjacent three-level features to provide a supplementary feature representation for the current-level features, effectively integrating the full context semantic information. Finally, we propose a context aggregation refinement module (CARM) to efficiently aggregate and refine the cross-level features to obtain clear prediction maps. Our extensive experiments on three benchmark datasets show that the MFNet model is an effective COD model and outperforms other state-of-the-art models in all four evaluation metrics ($S_{\alpha }$, $E_{\varphi }$, $F_{\beta }^w$, and $MAE$).

Infrared small target detection technology has attracted great attention and made significant research progress in military and civilian fields. However, most existing infrared small target detection algorithms fail to balance high detection accuracy and low computational complexity. To address this issue, we propose a simple and effective deep network framework for infrared small target detection. Our TNFPN architecture is a lightweight two-layer nested FPN structure. Specifically, each layer or stage in our FPN network embeds a small ResFPN network. The designed ResFPN block can capture richer contextual information from different scales, thereby learning finer feature representations. Meanwhile, the use of downsampling in ResFPN blocks enhances the depth of the structure without significantly increasing the computational complexity of the overall structure. Additionally, we design an Asymmetric Attention Interaction Module (AAIM), which filters out necessary semantic information contained in low-level features by using high-level features as weights, thus eliminating the negative impact of background noise on overall detection performance. We conduct experiments on three publicly available infrared small target datasets, and the results show that our proposed TNFPN, with only 1.85M parameters, achieves high detection accuracy while maintaining an inference speed of 41.43 FPS. The code is uploaded to: github.com/WoodXXX/TNFPN.

Satellite network resource management and scheduling technology are significant to constructing integrated information networks in heaven and earth. The difficulty in realizing this technology lies in improving resource utilization efficiency while ensuring the service quality of satellites and efficiently coordinating complex satellite network systems and services. This paper proposes a model, A Dynamic task scheduling method based on a UNified resource Management architecture(DUNM), based on the designed resource management architecture supported by dynamic scheduling algorithms to address the problems of low resource utilization, resource allocation, and task completion rate. First, with sufficient resources, the task execution time to complete a task is calculated based on the number of resources, task transmission time, task waiting time, etc. Secondly, based on the tasks assigned to satellites, the execution time of all functions with different transmission rates of communication links between satellites is calculated, and the total sum of all time consumption is analyzed. Finally, after simulation experiments and comparison with various baseline algorithms, about a 40% reduction in time to complete scheduled tasks and an almost 25% reduction in the average cost to finish a scheduling task, our method has higher scheduling efficiency and lower task completion revenue. It also guarantees a higher task completion rate while completing the tasks. Our approach attained a nearly 100% completion rate for scheduling tasks, which means that our algorithm can achieve the scheduling tasks faster and at high task revenue, thus improving the efficiency and economic efficiency of the whole system. Therefore, it validates the advantages of our method, such as high efficiency and high revenue.