Visible (VIS) and infrared (IR) image fusion (VIF) is a technique used to synthesize the fused of high visual perception. Existing methods typically work by discovering commons underlying two modalities fusing them in common space. However, these often ignore modality differences such as fuzzy details IR their well-designed architectures also lead slow speed. To address issues, we propose real-time end-to-end VIF model based on layer decomposition re-parameterization (LDRepFM). This composed...

Unsupervised domain adaptation (UDA) person re-identification (ReID) faces enormous challenges due to the severe shift between source and target domains, as well dramatic variations within domain. In this paper, address these issues, we propose a multi-loss gap minimization learning (MGML) approach for UDA ReID. Firstly, introduce part model learn discriminative patch features design Patch-based Part Ignoring (PPI) loss select reliable instances efficient of model. Then, given that typically...

Unsupervised person re-identification (Re-ID) targets to learn discriminative representations without annotations. Recently, clustering-based methods have shown promising performance, which utilize clustering generate identity pseudo labels for model optimization. Large intra-class variance mainly caused by domain discrepancy among cameras could lead noisy results. However, abundant camera-aware sample pairs relations not been exploited fully facilitate learning of features with...

<title>Abstract</title> Deep multi-view clustering (DMVC) aims to utilize the consistency of data learn a consensus representation using deep learning-based methods. However, existing methods overlook presence both semantic feature and topological structure information in data. Also, importance these two varies for heterogeneous To address issues, we propose Dual-Information Driven Multi-View Clustering Heterogeneous Data (DID-DMVC). Firstly, capture information, design Extractor (DIE),...

<title>Abstract</title> Image registration and fusion aim to align multi-modality images, generating fused image with richer information higher quality.Existing methods correct spatial misalignment through geometric transformation, semantic guidance, cross-modality complementarity.However, these adopt fixed receptive fields, overlooking local fine-grained features, which leads structural distortions edge artifacts.To address issues, this paper proposes an improved approach for misaligned via...

Unsupervised person re-identification (Re-ID) aims to learn semantic representations for retrieval without using identity labels. Most existing methods generate fine-grained patch features reduce noise in global feature clustering. However, these often compromise the discriminative structure and overlook consistency between features. To address problems, we propose a Person Intrinsic Semantic Learning (PISL) framework with diffusion model unsupervised Re-ID. First, design Spatial Diffusion...

Weakly supervised person re-identification (Re-ID) is appealing to handle real-world tasks by using state information that available without manual annotation. At present, most methods perform unsupervised cross domain (UCD) learning transferring the knowledge from labeled source unlabeled target domain, which results in poor performance due severe shift. To address this problem, paper, we utilize tracklet and camera as weak supervision propose a distribution discrepancy minimization (DDML)...

Instance grasping is a challenging robotic task when robot aims to grasp specified target object in cluttered scenes. In this paper, we propose novel end-to-end instance method using only monocular workspace and query images, where the image includes several objects contains object. To effectively extract discriminative features facilitate training process, learning-based method, referred as Constraint Co-Attention Network (CCAN), proposed which consists of constraint co-attention module...