Accurately predicting human future trajectories remains a core challenge in autonomous driving and human-robot interaction, primarily due to the inherent uncertainty of human motion and real-time requirements of decision-making. While generative models have shown promise in multimodal trajectory prediction, they often fail to simultaneously achieve high fidelity, high diversity, and fast sampling efficiency. To address this, we propose a unified trajectory prediction framework based on goal-guided truncated flow matching, which jointly optimizes these objectives. Specifically, we introduce a truncated flow matching strategy that initializes the denoising process from a context-aware intermediate state closer to the true trajectory distribution, reducing the number of sampling steps from 100 to just 2 and significantly enhancing inference efficiency. To enhance intent awareness and behavioral plausibility, we design an internal intention module that incorporates goal-guided cross-attention and is supervised by both destination regression and physical feasibility constraints, leading to more accurate and realistic predictions. Furthermore, to promote behavioral diversity and mitigate mode collapse, we propose a modality-aware diversity modeling mechanism that explicitly disentangles semantic differences among trajectory modes. Meanwhile, we introduce a diversity metric to quantitatively evaluate the diversity of the generated predictions. Extensive experiments on three public datasets show that our method achieves competitive accuracy, significantly enhances motion diversity, and delivers superior inference efficiency compared to existing models. These results highlight its strong potential for real-world deployment in safety-critical applications.

Defocus deblurring is a challenging task due to the spatially varying blur. Recent works have shown impressive results in data-driven approaches using dual-pixel (DP) sensors. Quad-pixel (QP) sensors represent an advanced evolution of DP sensors, providing four distinct sub-aperture views in contrast to only two views offered by DP sensors. However, research on QP-based defocus deblurring is scarce. In this paper, we propose a novel end-to-end learning-based approach for defocus deblurring that leverages QP data. To achieve this, we design a QP defocus and all-in-focus image pair acquisition method and provide a QP Defocus Deblurring (QPDD) dataset containing 4,935 image pairs. We then introduce a Local-gate assisted Mamba Network (LMNet), which includes a two-branch encoder and a Simple Fusion Module (SFM) to fully utilize features of sub-aperture views. In particular, our LMNet incorporates a Local-gate assisted Mamba Block (LAMB) that mitigates local pixel forgetting and channel redundancy within Mamba, and effectively captures global and local dependencies. By extending the defocus deblurring task from a DP-based to a QP-based approach, we demonstrate significant improvements in restoring sharp images. Comprehensive experimental evaluations further indicate that our approach outperforms state-of-the-art methods.

Quad Photodiode (QPD) sensors represent an evolution by providing four sub-views, whereas dual-pixel (DP) sensors are limited to two sub-views. In addition to enhancing auto-focus performance, QPD sensors also enable disparity estimation in horizontal and vertical directions. However, the characteristics of QPD sensors, including uneven illumination across sub-views and the narrow baseline, render algorithm design difficult. Furthermore, effectively utilizing the two-directional disparity of QPD sensors remains a challenge. The scarcity of QPD disparity datasets also limits the development of learning-based methods. In this work, we address these challenges by first proposing a DP-Net for DP disparity estimation. Specifically, we design an illumination-invariant module to reduce the impact of illumination, followed by a coarse-to-fine module to estimate sub-pixel disparity. Building upon the DPNet, we further propose a QuadNet, which integrates the two-directional disparity via an edge-aware fusion module. To facilitate the evaluation of our approaches, we propose the first QPD disparity dataset QPD2K, comprising 2,100 real-world QPD images and corresponding disparity maps. Experiments demonstrate that our approaches achieve state-of-the-art performance in DP and QPD disparity estimation.

Fire detection technology based on deep learning methods has become a prevalent practice. However, the performance of current YOLO-based detection models is limited by NMS, and DETR-based detection models struggle with real-time performance. To address these challenges, a new fire detection model, FasterGold-DETR, is proposed. Firstly, this model introduces an innovative backbone network, FasterRepNet, which efficiently captures and retains feature information, thereby accelerating the model’s convergence speed. Secondly, we propose the AIFI-GD hybrid encoder to reduce information loss in intra-scale and cross-scale feature interactions and improve the ability to detect fire of different sizes. Furthermore, to adapt the complex fire scenarios, we extend the dataset based on the KMU Fire and Smoke database and replace the loss function with WIoU to enhance the model’s robustness. Experiments show that our proposed model outperforms mainstream object detection models in terms of accuracy and complexity.

Deep learning is widely used in the early diagnosis of Alzheimer ’s disease (AD) in recent years, and has achieved better performance than traditional machine learning methods. The current deep methods for early diagnosis of AD mainly adopt a fully supervised method, which relies too much on a large number of labeled high-quality medical image data, and has a large performance loss in under-labeled data scenarios. Meanwhile,the existing semi-supervised methods ignore the reliability of pseudo-labeling. Aiming at the above problems, a Dual Augmentation Semi-supervised Learning (DASSL) method is proposed. DASSL designs a feature enhancement method based on attention mechanism and combines it with a spatial enhancement method more suitable for medical images, which provides a new solution for early feature recognition of Alzheimer ’s disease. At the same time, a new loss function is designed, and combined with the threshold, high confidence samples are selected to cope with the challenge of high sample impurity rate. The DASSL method proposed in this paper was evaluated in 518 subjects on the ADNI-1 dataset. The experimental results show that DASSL achieves the best accuracy and stability in classifying AD, MCI, and NC (more than 95% accuracy for classifying AD and CN task, and more than 90% accuracy for classifying MCI and CN) compared to other semi-supervised methods.

Abstract Siamese network based the tracker is a hot topic in the field of visual object tracking. However, Siamese trackers still have a robustness gap compared with state‐of‐the‐art algorithms. Therefore, focusing on the issue, this letter adds Frequency Channel Attention (FCA) and adaptive template feature map to the framework of Siamese neural network. FCA can enhance feature representation of effective channels and improve feature discrimination by modeling the correlation between each channel of the image. In this algorithm, by theoretical analysis and experimental validation, restriction is broken through a simple yet effective FCA network sampling strategy and a Siamese‐FCA tracker with significant performance gain is successfully trained. Meanwhile, in order to better adjust the proportion between target and background, the tracker selects suitable size of the target feature map. Moreover, extensive ablation studies are conducted to demonstrate the effectiveness of the proposed tracker. Fairly, the experimental results of five test benchmarks, including OTB2013, OTB2015, VOT2016, VOT2018 and UAV123 datasets, shows that the proposed algorithm performs outstanding. In particular, the issue of similarity and small target tracking failure is overcome. The average running frame rate reaches 86 frames per second, which can meet the real‐time requirements.

The head-mounted display (HMD) for 360-degree videos cannot be shared by multiple users, which significantly increases the bandwidth consumption when multiple users request the same video content simultaneously. This paper proposes a server-side hybrid multicast-unicast strategy for multi-user adaptive 360-degree video streaming, aiming to ensure the overall quality of experience (QoE) of the users in a bandwidth-constrained environment. A framework is established to realise this strategy, through which the delivery mode (multicast or unicast) and the corresponding bitrate for each tile can be jointly adapted to the dynamics of users' throughputs and field of views (FoVs). To maximize the overall QoE, a user clustering method is proposed to classify the users into several multicast clusters. Based on these clusters, we then formulate the joint delivery mode selection and rate adaptation problem as a non-linear integer programming problem which can be solved using the steepest ascent gradient algorithm. Finally, experiments are carried out to verify the performance of the proposed strategy.

How to balance the tradeoff between the user experience and bandwidth utilization emerges a critical challenge for multi-user 360-degree video adaptive streaming. This paper studies the server-side rate adaptation strategy for multiple users which are competing for the server bandwidth capacity. A tile visibility probability model is established, by which the tiles are classified into predicted, marginal and invisible types. A fine-grained rate adaptation problem is formulated as a nonlinear integer programming (NIP) problem, which aims at maximizing the video quality and navigation smoothness for multiple users. Thereafter, a steepest ascent algorithm with feasible starting point is developed to solve the proposed NIP problem in polynomial time. Finally, simulation results verify the performance of the proposed rate adaptation strategy.

In this paper, we propose a new image interpolation algorithm by using geometric subdivision. Similar to image upsampling, the geometric subdivision can supplement unknown data according to a certain rule, but it can only generate smooth data. To preserve the sharp edges of the high-resolution image, we adopt a rational subdivision scheme. By adjusting the weight coefficients of the rational subdivision, we can control the mesh shape near sharp edges. Hence, the image edges are preserved.

Image retrieval methods have been dramatically developed in the last decade. In this paper, we propose a novel method for image retrieval based on manifold structures of canonical images. Firstly, we present the image normalization process to find a set of canonical images that anchors the probabilistic distributions around the real data manifolds to learn the representations that better encode the manifold structures in general high-dimensional image space. In addition, we employ the canonical images as the centers of the conditional multivariate Gaussian distributions. This approach allows to learn more detailed structures of the partial manifolds resulting in improved representation of the high level properties of scene images. Furthermore, we use the probabilistic framework of the extended model to retrieve images based on the similarity measure of reciprocal likelihood of pairs of images and the sum of likelihood of one of two images based on the other’s best distributions. We estimate our method using SUN database. In the experiments on scene image retrieval, the proposed method is efficient, and exhibits superior capabilities compared to other methods, such as GIST.