Electricity price forecasting remains a critical research focus in modern power systems, where future electricity prices are influenced not only by historical electricity price patterns but also by exogenous factors. Traditional approaches often employ single-stream recurrent neural networks to integrate exogenous variables with historical electricity price sequences. However, since historical electricity prices and exogenous information inherently represent distinct modalities, single-stream architectures are suboptimal for processing such multimodal data, leading to inefficient feature extraction. To address this limitation, we propose DSFormer, a dual-stream transformer framework designed for electricity price forecasting. This dual-stream architecture employs parallel processing, with the parsing stream handling historical electricity price sequences and the exogenous stream processing exogenous variables. Specifically, DSFormer comprises an encoder and a decoder. The encoder incorporates two key components: a fusion mutual attention mechanism and a multispace feedforward network. The fusion mutual attention mechanism comprises a pair of mutual attention layers, followed by a fusion attention layer. The mutual attention layers independently inject historical electricity price information into the exogenous stream and exogenous information into the parsing stream, while the fusion attention layer further refines the features from dual streams. The multispace feedforward network projects the features from dual streams into subspaces and performs cross-subspace interactions based on the gating mechanism, thereby enhancing representation capabilities while achieving enhanced computational efficiency. Finally, the decoder infers future electricity prices from these refined features. Extensive experimental results demonstrate that DSFormer outperforms existing state-of-the-art methods in terms of prediction accuracy while maintaining lower computational complexity.

Vehicle-infrastructure collaborative perception has demonstrated significant potential in extending perception range and enhancing 3D object detection accuracy. However, there remain challenges due to limitations in communication bandwidth and the redundancy of shared information. To address these issues, we propose Why2comm, a novel framework for vehicle-infrastructure collaborative perception. By incorporating driving trajectories and region decoupling, Why2comm focuses information sharing on key areas that directly impact driving tasks, thereby minimizing redundant data transmission and improving collaborative perception efficiency. Specifically, we design a dynamic trajectory-aware message filtering module that investigates the potential correlations between driving trajectories and instance features, utilizing an attention mechanism to identify and extract the critical instance features. Furthermore, we propose a region-aware feature transmission module that decouples shared and exclusive regions, achieving efficient data transmission. Finally, we develop a position-guided fusion module to effectively integrate and enhance features from both vehicle and infrastructure. Evaluation results on the V2XSet, DAIR-V2X, and DeepAccident datasets show that Why2comm achieves superior detection performance compared to state-of-the-art methods while requiring merely 10.7% of the communication bandwidth.

Trajectory prediction is a continuing concern within autonomous vehicles. Psychological research shows that pedestrian traveling is a cyclic alternation. Pedestrians constantly interact with their surroundings, including social agents and physical environments, and plan paths to achieve goals. Nevertheless, most existing trajectory prediction methods are based on a single-stage design, which runs counter to traffic psychology principles. In this work, we present MPNet, a novel multi-stage progressive convolutional neural network that decomposes complicated trajectory prediction into multiple manageable components, where lightweight sub-networks handle each stage with a divide-and-conquer methodology. Specifically, our sub-networks are based on the encoder-decoder architecture, in which we capture interactive information and estimate goals to achieve trajectory prediction. The communication among sub-networks depends on a novel cross-stage fusion design. We introduce a feedback channel mutual attention mechanism and a cross-sub-network fusion unit to enable efficient information sharing across different stages. At each stage, we develop a symmetric gated supervision module to supervise future trajectory generation from coarse to fine. Extensive experiments demonstrate that MPNet achieves state-of-the-art performance, reducing Average Displacement Error (ADE) and Final Displacement Error (FDE) by 5.6%/7.4% on the ETH and UCY Datasets, 7.5%/7.7% on the Stanford Drone Dataset, and 7.1%/22.5% on the Intersection Drone Dataset, while maintaining comparable computational complexity. Additionally, our MPNet-Tiny variant reduces parameters by 90.5% and inference time by 1.83 seconds with competitive accuracy.

Incorporating both RGB and depth images has proven effective for enhancing the performance of semantic segmentation. However, current RGB-D semantic segmentation methods tend to overlook the critical role of cross-modal difference information during fusion, leading to the undesired suppression of discriminative cues and a failure to achieve potent cross-modal complementary fusion. In this article, a novel RGB-D semantic segmentation approach that realizes the efficient utilization of multimodal information is proposed. To address the issue of the suppression of cross-modal difference information, we propose a dynamic frequency-spatial difference-aware fusion module adept at explicitly emphasizing cross-modal differences, capturing vital features in the frequency domain, and using them to aggregate spatial context information of multimodal features. We also present a novel soft-edge loss to meticulously handle complex scenes by supervising different regions respectively. In addition, a progressive calibration context module is designed to enhance global contextual information by capturing multiscale multimodal representations. Extensive experiments on two public RGB-D datasets demonstrate that the proposed DFNet achieves highly competitive performance compared to state-of-the-art methods, making it well-suited for assisting indoor robots.

Diffusion probabilistic models (DPMs) have recently achieved brilliant achievements in computer vision. Inspired by the success of DPMs, we present TrajDiff, a model based on conditional diffusion probabilistic models for agent future trajectory prediction, which speculates the agent future states through a series of stochastic iterative denoising processes. Specifically, we map the trajectory prediction task into the latent heatmap space, translating hard keypoint prediction into soft cluster center learning. The core architecture is a U-shaped encoder-decoder network (U-Net) that is trained with a denoising objective. During inference, conditioned on the observed past trajectory heatmaps, random pure Gaussian noise is initialized to drive the reverse sampling process. The U-Net iteratively removes various levels of Gaussian noise from initialized images, resembling Langevin dynamics, and generates multi-modal predicted future trajectory heatmaps. Furthermore, we introduce a novel residual block with a mutual attention mechanism that can elegantly consider the interactions between the agent and the surrounding environment at multiple scales, assisting in generating physically and socially acceptable trajectories. We verify TrajDiff on the Stanford Drone Dataset and the ETH and UCY Datasets. The experimental results show that TrajDiff outperforms previous state-of-the-art methods with considerable accuracy gains, while significantly reducing computational requirements.

Deep feature fusion plays a significant role in the strong learning ability of convolutional neural networks (CNNs) for computer vision tasks. Recently, works continually demonstrate the advantages of efficient aggregation strategy and some of them refer to multiscale representations. In this article, we describe a novel network architecture for high-level computer vision tasks where densely connected feature fusion provides multiscale representations for the residual network. We term our method the ResDNet which is a simple and efficient backbone made up of sequential ResDNet modules containing the variants of dense blocks named sliding dense blocks (SDBs). Compared with DenseNet, ResDNet enhances the feature fusion and reduces the redundancy by shallower densely connected architectures. Experimental results on three classification benchmarks including CIFAR-10, CIFAR-100, and ImageNet demonstrate the effectiveness of ResDNet. ResDNet always outperforms DenseNet using much less computation on CIFAR-100. On ImageNet, ResDNet-B-129 achieves 1.94% and 0.89% top-1 accuracy improvement over ResNet-50 and DenseNet-201 with similar complexity. Besides, ResDNet with more than 1000 layers achieves remarkable accuracy on CIFAR compared with other state-of-the-art results. Based on MMdetection implementation of RetinaNet, ResDNet-B-129 improves mAP from 36.3 to 39.5 compared with ResNet-50 on COCO dataset.

Using light-weight architectures or reasoning on low-resolution images, recent methods realize very fast scene parsing, even running at more than 100 FPS on a single GPU. However, there is still a significant gap in performance between these real-time methods and the models based on dilation backbones. To this end, we proposed a family of deep dual-resolution networks (DDRNets) for real-time and accurate semantic segmentation, which consist of deep dual-resolution backbones and enhanced low-resolution contextual information extractors. The two deep branches and multiple bilateral fusions of backbones generate higher quality details compared to existing two-pathway methods. The enhanced contextual information extractor named Deep Aggregation Pyramid Pooling Module (DAPPM) enlarges effective receptive fields and fuses multi-scale context based on low-resolution feature maps with little time cost. Our method achieves a new state-of-the-art trade-off between accuracy and speed on both Cityscapes and CamVid dataset. For the input of full resolution, on a single 2080Ti GPU without hardware acceleration, DDRNet-23-slim yields 77.4% mIoU at 102 FPS on Cityscapes test set and 74.7% mIoU at 230 FPS on CamVid test set. With widely used test augmentation, our method is superior to most state-of-the-art models and requires much less computation. Codes and trained models are available athttps://github.com/ydhongHIT/DDRNet.