Person re-identification (re-ID) tackles the problem of matching person images with same identity from different cameras. In practical applications, due to differences in camera performance and distance between cameras persons interest, captured usually have various resolutions. This problem, named Cross-Resolution Re-identification, presents a great challenge for accurate matching. this paper, we propose Deep High-Resolution Pseudo-Siamese Framework (PS-HRNet) solve above problem....

The performance of person re-identification (re-ID) is easily affected by illumination variations caused different shooting times, places and cameras. Existing illumination-adaptive methods usually require annotating cross-camera pedestrians on each scale, which unaffordable for a long-term retrieval system. cross-illumination problem presents great challenge accurate matching. In this paper, we propose novel method to tackle task, only needs annotate one scale. Specifically, (i)...

Person re-identification (re-ID) aims to match the same person across different cameras. However, most existing re-ID methods assume that people wear clothes in views, which limit their performance identifying target pedestrians who change clothes. Cloth-changing is a quite challenging problem as occupying large number of pixels an image becomes invalid or even misleads information. To tackle this problem, we propose novel Multi-biometric Unified Network (MBUNet) for learning robustness...

Object detection is dedicated to finding objects in an image and estimate their categories locations. Recently, object algorithms suffer from a loss of semantic information the deeper feature maps due deepening backbone network. For example, when using complex networks, existing fusion methods cannot fuse different layers effectively. In addition, anchor-free fail accurately predict same learning mechanisms regression centrality prediction branches. To address above problem, we propose...

Studies on human motion have attracted a lot of attentions. Human capture data, which much more precisely records than videos do, has been widely used in many areas. Motion segmentation is an indispensable step for related applications, but current methods data do not effectively model some important characteristics such as Riemannian manifold structure and containing non-Gaussian noise. In this paper, we convert the into temporal subspace clustering problem. Under framework sparse...

The goal of unsupervised person re-identification (Re-ID) is to use unlabeled images learn discriminative features. In recent years, many approaches have adopted clustered pseudo labels construct proxies for contrastive learning, and thereby achieved great success. However, existing methods this kind only utilize local structures within IDs design their while ignoring the relations between samples different IDs, which limits improvement inter-ID ability. To resolve issue, we propose a Global...

Human motion capture data has been widely used in many areas, but it involves a complex process and the captured inevitably contains missing due to occlusions caused by actor's body or clothing. Motion recovery, which aims recover underlying complete sequence from its degraded observation, still remains as challenging task nonlinear structure kinematics property embedded data. Low-rank matrix completion-based methods have shown promising performance short-time-missing recovery problems....

Adverse weather conditions such as haze and snowfall can degrade the quality of captured images affect performance drone detection. Therefore, it is challenging to locate identify targets in adverse scenarios. In this paper, a novel model called Object Detection Foggy Condition with YOLO (ODFC-YOLO) proposed, which performs image dehazing object detection jointly by multi-task learning approach. Our consists subnet subnet, be trained end-to-end optimize both tasks. Specifically, we propose...

Dictionary learning has produced state-of-the-art results in various classification tasks. However, if the training data have a different distribution than testing data, learned sparse representation might not be optimal. Recently, several domain-adaptive dictionary (DADL) methods and kernels been proposed achieved impressive performance. performance of these single kernel-based heavily depends on choice kernel, question how to combine multiple kernel (MKL) with DADL framework well studied....

As a prevailing task in video surveillance and forensics field, person re-identification (re-ID) aims to match images captured from non-overlapped cameras. In unconstrained scenarios, often suffer the resolution mismatch problem, i.e., Cross-Resolution Person Re-ID. To overcome this most existing methods restore low (LR) high (HR) by super-resolution (SR). However, they only focus on HR feature extraction ignore valid information original LR images. work, we explore influence of resolutions...

Confucianism, Buddhism and Taoism are three main classic Chinese philosophy schools, which all deal with the question of how one should live. In this paper, we first review these ancient recommendations next consider whether they promise a happy life in present-day society. Recommended behaviours found texts compared conditions for happiness as observed empirical investigations. Classic Confucianism appears to offer most apt advice finding society, particular because it recommends that be...

Recently, sparse representation-based classification (SRC) has been widely studied and produced state-of-the-art results in various tasks. Learning useful computationally convenient representations from complex redundant highly variable visual data is crucial for the success of SRC. However, how to find best feature representation work with SRC remains an open question. In this paper, we present a novel discriminative projection learning approach objective seeking matrix such that learned...

In contemporary times, owing to the swift advancement of Unmanned Aerial Vehicles (UAVs), there is enormous potential for use UAVs ensure public safety. Most research on capturing images by mainly focuses object detection and tracking tasks, but few studies have focused UAV re-identification task. addition, in real-world scenarios, objects frequently get together groups. Therefore, re-identifying groups poses a significant challenge. this paper, novel dynamic screening strategy based feature...

Sparse representation-based classification (SRC) has been developed and shown great potential for real-world application. Based on SRC, Yang et al. devised an SRC steered discriminative projection (SRC-DP) method. However, as a linear algorithm, SRC-DP cannot handle the data with highly nonlinear distribution. Kernel sparse classifier (KSRC) is non-linear extension of can remedy drawback SRC. KSRC requires use predetermined kernel function selection its parameters difficult. Recently,...