Learning Inverse Laplacian Pyramid for Progressive Depth Completion

Depth completion endeavors to reconstruct a dense depth map from sparse depth measurements, leveraging the information provided by a corresponding color image. Existing approaches mostly hinge on single-scale propagation strategies that iteratively ameliorate initial coarse depth estimates through pixel-level message passing. Despite their commendable outcomes, these techniques are frequently hampered by computational inefficiencies and a limited grasp of scene context. To circumvent these challenges, we introduce LP-Net, an innovative framework that implements a multi-scale, progressive prediction paradigm based on Laplacian Pyramid decomposition. Diverging from propagation-based approaches, LP-Net initiates with a rudimentary, low-resolution depth prediction to encapsulate the global scene context, subsequently refining this through successive upsampling and the reinstatement of high-frequency details at incremental scales. We have developed two novel modules to bolster this strategy: 1) the Multi-path Feature Pyramid module, which segregates feature maps into discrete pathways, employing multi-scale transformations to amalgamate comprehensive spatial information, and 2) the Selective Depth Filtering module, which dynamically learns to apply both smoothness and sharpness filters to judiciously mitigate noise while accentuating intricate details. By integrating these advancements, LP-Net not only secures state-of-the-art (SOTA) performance across both outdoor and indoor benchmarks such as KITTI, NYUv2, and TOFDC, but also demonstrates superior computational efficiency. At the time of submission, LP-Net ranks 1st among all peer-reviewed methods on the official KITTI leaderboard.

Decision Transformers for RIS-Assisted Systems with Diffusion Model-Based Channel Acquisition

Reconfigurable intelligent surfaces (RISs) have been recognized as a revolutionary technology for future wireless networks. However, RIS-assisted communications have to continuously tune phase-shifts relying on accurate channel state information (CSI) that is generally difficult to obtain due to the large number of RIS channels. The joint design of CSI acquisition and subsection RIS phase-shifts remains a significant challenge in dynamic environments. In this paper, we propose a diffusion-enhanced decision Transformer (DEDT) framework consisting of a diffusion model (DM) designed for efficient CSI acquisition and a decision Transformer (DT) utilized for phase-shift optimizations. Specifically, we first propose a novel DM mechanism, i.e., conditional imputation based on denoising diffusion probabilistic model, for rapidly acquiring real-time full CSI by exploiting the spatial correlations inherent in wireless channels. Then, we optimize beamforming schemes based on the DT architecture, which pre-trains on historical environments to establish a robust policy model. Next, we incorporate a fine-tuning mechanism to ensure rapid beamforming adaptation to new environments, eliminating the retraining process that is imperative in conventional reinforcement learning (RL) methods. Simulation results demonstrate that DEDT can enhance efficiency and adaptability of RIS-aided communications with fluctuating channel conditions compared to state-of-the-art RL methods.

Dual Scale-aware Adaptive Masked Knowledge Distillation for Object Detection

Recent feature masking knowledge distillation methods make use of attention mechanisms to identify either important spatial regions or channel clues for discriminative feature reconstruction. However, most of existing strategies perform global attention-guided feature masking distillation without delving into fine-grained visual clues in feature maps. In particular, uncovering locality-aware clues across different scales are conducive to reconstructing region-aware features, thereby significantly benefiting distillation performance. In this study, we propose a fine-grained adaptive feature masking distillation framework for accurate object detection. Different from previous methods in which global masking is performed on single-scale feature maps, we explore the scale-aware feature masking by performing feature distillation across various scales, such that the object-aware locality is encoded for improved feature reconstruction. In addition, our fine-grained feature distillation strategy is combined with a masking logits distillation scheme in which logits difference between teacher and student networks is utilized to guide the distillation process. Thus, it can help the student model to better learn from the teacher counterpart with improved knowledge transfer. Extensive experiments for detection task demonstrate the superiority of our method. For example, when RetinaNet, RepPoints and Cascade Mask RCNN are used as teacher detectors, the student network achieves mAP scores of 41.5\%, 42.9\%, and 42.6\%, respectively, outperforming state-of-the-art methods such as DMKD and FreeKD.

SFADNet: Spatio-temporal Fused Graph based on Attention Decoupling Network for Traffic Prediction

In recent years, traffic flow prediction has played a crucial role in the management of intelligent transportation systems. However, traditional prediction methods are often limited by static spatial modeling, making it difficult to accurately capture the dynamic and complex relationships between time and space, thereby affecting prediction accuracy. This paper proposes an innovative traffic flow prediction network, SFADNet, which categorizes traffic flow into multiple traffic patterns based on temporal and spatial feature matrices. For each pattern, we construct an independent adaptive spatio-temporal fusion graph based on a cross-attention mechanism, employing residual graph convolution modules and time series modules to better capture dynamic spatio-temporal relationships under different fine-grained traffic patterns. Extensive experimental results demonstrate that SFADNet outperforms current state-of-the-art baselines across four large-scale datasets.

Unleashing Correlation and Continuity for Hyperspectral Reconstruction from RGB Images

Reconstructing Hyperspectral Images (HSI) from RGB images can yield high spatial resolution HSI at a lower cost, demonstrating significant application potential. This paper reveals that local correlation and global continuity of the spectral characteristics are crucial for HSI reconstruction tasks. Therefore, we fully explore these inter-spectral relationships and propose a Correlation and Continuity Network (CCNet) for HSI reconstruction from RGB images. For the correlation of local spectrum, we introduce the Group-wise Spectral Correlation Modeling (GrSCM) module, which efficiently establishes spectral band similarity within a localized range. For the continuity of global spectrum, we design the Neighborhood-wise Spectral Continuity Modeling (NeSCM) module, which employs memory units to recursively model the progressive variation characteristics at the global level. In order to explore the inherent complementarity of these two modules, we design the Patch-wise Adaptive Fusion (PAF) module to efficiently integrate global continuity features into the spectral features in a patch-wise adaptive manner. These innovations enhance the quality of reconstructed HSI. We perform comprehensive comparison and ablation experiments on the mainstream datasets NTIRE2022 and NTIRE2020 for the spectral reconstruction task. Compared to the current advanced spectral reconstruction algorithms, our designed algorithm achieves State-Of-The-Art (SOTA) performance.

Completion as Enhancement: A Degradation-Aware Selective Image Guided Network for Depth Completion

In this paper, we introduce the Selective Image Guided Network (SigNet), a novel degradation-aware framework that transforms depth completion into depth enhancement for the first time. Moving beyond direct completion using convolutional neural networks (CNNs), SigNet initially densifies sparse depth data through non-CNN densification tools to obtain coarse yet dense depth. This approach eliminates the mismatch and ambiguity caused by direct convolution over irregularly sampled sparse data. Subsequently, SigNet redefines completion as enhancement, establishing a self-supervised degradation bridge between the coarse depth and the targeted dense depth for effective RGB-D fusion. To achieve this, SigNet leverages the implicit degradation to adaptively select high-frequency components (e.g., edges) of RGB data to compensate for the coarse depth. This degradation is further integrated into a multi-modal conditional Mamba, dynamically generating the state parameters to enable efficient global high-frequency information interaction. We conduct extensive experiments on the NYUv2, DIML, SUN RGBD, and TOFDC datasets, demonstrating the state-of-the-art (SOTA) performance of SigNet.

Multimodal joint prediction of traffic spatial-temporal data with graph sparse attention mechanism and bidirectional temporal convolutional network

Traffic flow prediction plays a crucial role in the management and operation of urban transportation systems. While extensive research has been conducted on predictions for individual transportation modes, there is relatively limited research on joint prediction across different transportation modes. Furthermore, existing multimodal traffic joint modeling methods often lack flexibility in spatial-temporal feature extraction. To address these issues, we propose a method called Graph Sparse Attention Mechanism with Bidirectional Temporal Convolutional Network (GSABT) for multimodal traffic spatial-temporal joint prediction. First, we use a multimodal graph multiplied by self-attention weights to capture spatial local features, and then employ the Top-U sparse attention mechanism to obtain spatial global features. Second, we utilize a bidirectional temporal convolutional network to enhance the temporal feature correlation between the output and input data, and extract inter-modal and intra-modal temporal features through the share-unique module. Finally, we have designed a multimodal joint prediction framework that can be flexibly extended to both spatial and temporal dimensions. Extensive experiments conducted on three real datasets indicate that the proposed model consistently achieves state-of-the-art predictive performance.

Multi-View Fusion Neural Network for Traffic Demand Prediction

The extraction of spatial-temporal features is a crucial research in transportation studies, and current studies typically use a unified temporal modeling mechanism and fixed spatial graph for this purpose. However, the fixed spatial graph restricts the extraction of spatial features for similar but not directly connected nodes, while the unified temporal modeling mechanism overlooks the heterogeneity of temporal variation of different nodes. To address these challenges, a multi-view fusion neural network (MVFN) approach is proposed. In this approach, spatial local features are extracted through the use of a graph convolutional network (GCN), and spatial global features are extracted using a cosine re-weighting linear attention mechanism (CLA). The GCN and CLA are combined to create a graph-cosine module (GCM) for the extraction of overall spatial features. Additionally, the multi-channel separable temporal convolutional network (MSTCN) makes use of a multi-channel temporal convolutional network (MTCN) at each layer to extract unified temporal features, and a separable temporal convolutional network (STCN) to extract independent temporal features. Finally, the spatial-temporal feature data is input into the prediction layer to obtain the final result. The model has been validated on two traffic demand datasets and achieved the best prediction accuracy.

Guided Real Image Dehazing using YCbCr Color Space

Image dehazing, particularly with learning-based methods, has gained significant attention due to its importance in real-world applications. However, relying solely on the RGB color space often fall short, frequently leaving residual haze. This arises from two main issues: the difficulty in obtaining clear textural features from hazy RGB images and the complexity of acquiring real haze/clean image pairs outside controlled environments like smoke-filled scenes. To address these issues, we first propose a novel Structure Guided Dehazing Network (SGDN) that leverages the superior structural properties of YCbCr features over RGB. It comprises two key modules: Bi-Color Guidance Bridge (BGB) and Color Enhancement Module (CEM). BGB integrates a phase integration module and an interactive attention module, utilizing the rich texture features of the YCbCr space to guide the RGB space, thereby recovering clearer features in both frequency and spatial domains. To maintain tonal consistency, CEM further enhances the color perception of RGB features by aggregating YCbCr channel information. Furthermore, for effective supervised learning, we introduce a Real-World Well-Aligned Haze (RW$^2$AH) dataset, which includes a diverse range of scenes from various geographical regions and climate conditions. Experimental results demonstrate that our method surpasses existing state-of-the-art methods across multiple real-world smoke/haze datasets. Code and Dataset: \textcolor{blue}{\url{https://github.com/fiwy0527/AAAI25_SGDN.}}

Exploiting Multimodal Spatial-temporal Patterns for Video Object Tracking

Multimodal tracking has garnered widespread attention as a result of its ability to effectively address the inherent limitations of traditional RGB tracking. However, existing multimodal trackers mainly focus on the fusion and enhancement of spatial features or merely leverage the sparse temporal relationships between video frames. These approaches do not fully exploit the temporal correlations in multimodal videos, making it difficult to capture the dynamic changes and motion information of targets in complex scenarios. To alleviate this problem, we propose a unified multimodal spatial-temporal tracking approach named STTrack. In contrast to previous paradigms that solely relied on updating reference information, we introduced a temporal state generator (TSG) that continuously generates a sequence of tokens containing multimodal temporal information. These temporal information tokens are used to guide the localization of the target in the next time state, establish long-range contextual relationships between video frames, and capture the temporal trajectory of the target. Furthermore, at the spatial level, we introduced the mamba fusion and background suppression interactive (BSI) modules. These modules establish a dual-stage mechanism for coordinating information interaction and fusion between modalities. Extensive comparisons on five benchmark datasets illustrate that STTrack achieves state-of-the-art performance across various multimodal tracking scenarios. Code is available at: https://github.com/NJU-PCALab/STTrack.