Person Re-identification (re-id) aims to match people across non-overlapping camera views in a public space. It is challenging problem because many captured surveillance videos wear similar clothes. Consequently, the differences their appearance are often subtle and only detectable at right location scales. Existing re-id models, particularly recently proposed deep learning based ones single scale. In contrast, this paper, novel multi-scale model proposed. Our able learn discriminative...

Person re-identification (ReID) is now an active research topic for AI-based video surveillance applications such as specific person search, but the practical issue that target person(s) may change clothes (clothes inconsistency problem) has been overlooked long. For first time, this paper systematically studies problem. We overcome difficulty of lack suitable dataset, by collecting a small yet representative real dataset testing whilst building large realistic synthetic training and deeper...

Person re-identification (re-id) aims to match people across non-overlapping camera views in a public space. This is challenging problem because the captured surveillance videos often wear similar clothing. Consequently, differences their appearance are typically subtle and only detectable at particular locations scales. In this paper, we propose deep re-id network (MuDeep) that composed of two novel types layers - multi-scale learning layer, leader-based attention layer. Specifically,...

To counter the outbreak of COVID-19, accurate diagnosis suspected cases plays a crucial role in timely quarantine, medical treatment, and preventing spread pandemic. Considering limited training resources (e.g, time budget), we propose Multi-task Multi-slice Deep Learning System (M <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> Lung-Sys) for multi-class lung pneumonia screening from CT imaging, which only consists two 2D CNN networks,...

Amodal object segmentation is a challenging task that involves segmenting both visible and occluded parts of an object. In this paper, we propose novel approach, called Coarse-to-Fine Segmentation (C2F-Seg), addresses problem by progressively modeling the amodal segmentation. C2F-Seg initially reduces learning space from pixel-level image to vector-quantized latent space. This enables us better handle long-range dependencies learn coarse-grained segment visual features segments. However,...

Lifelong Person Re-Identification (LReID) extends traditional ReID by requiring systems to continually learn from non-overlapping datasets across different times and locations, adapting new identities while preserving knowledge of previous ones. Existing approaches, either rehearsal-free or rehearsal-based, still suffer the problem catastrophic forgetting since they try cram diverse into one fixed model. To overcome this limitation, we introduce a novel framework AdalReID, that adopts...

Deep Convolutional Neural Networks (CNN) has achieved significant success in computer vision field. However, the high computational cost of deep complex models prevents deployment on edge devices with limited memory and resource. In this paper, we proposed a novel filter pruning for convolutional neural networks compression, namely spectral clustering soft self-adaption manners (SCSP). We first apply filters layer by to explore their intrinsic connections only count efficient groups. By...

Person Re-identification (re-id) aims to match people across non-overlapping camera views in a public space. It is challenging problem because many captured surveillance videos wear similar clothes. Consequently, the differences their appearance are often subtle and only detectable at right location scales. Existing re-id models, particularly recently proposed deep learning based ones single scale. In contrast, this paper, novel multi-scale model proposed. Our able learn discriminative...

It is challenging in learning a makeup-invariant face verification model, due to (1) insufficient makeup/non-makeup training pairs, (2) the lack of diverse makeup faces, and (3) significant appearance changes caused by cosmetics. To address these challenges, we propose unified Face Morphological Multi-branch Network (FMMu-Net) for verification, which can simultaneously synthesize many faces through morphology network (FM-Net) effectively learn cosmetics-robust representations using...

With the wide applications of deep neural network models in various computer vision tasks, more and works study model vulnerability to adversarial examples. For data-free black box attack scenario, existing methods are inspired by knowledge distillation, thus usually train a substitute learn from target using generated data as input. However, always has static structure, which limits ability for tasks. In this paper, we propose novel dynamic training method encourage better faster model....

Conventional 3D object detection approaches concentrate on bounding boxes representation learning with several parameters, i.e., localization, dimension, and orientation. Despite its popularity universality, such a straightforward paradigm is sensitive to slight numerical deviations, especially in localization. By exploiting the property that point clouds are naturally captured surface of objects along accurate location intensity information, we introduce new perspective views box regression...

We present a significant breakthrough in 3D shape generation by scaling it to unprecedented dimensions. Through the adaptation of Auto-Regressive model and utilization large language models, we have developed remarkable with an astounding 3.6 billion trainable parameters, establishing as largest date, named Argus-3D. Our approach addresses limitations existing methods enhancing quality diversity generated shapes. To tackle challenges high-resolution generation, our incorporates tri-plane...

Cloth-changing person Re-IDentification (Re-ID) aims at recognizing the same with clothing changes across non-overlapping cameras. Conventional Re-ID methods usually bias model's focus on cloth-related appearance features rather than identity-sensitive associated biological traits. Recently, advanced cloth-changing either resort to identity-related auxiliary modalities (e.g., sketches, silhouettes, keypoints and 3D shapes) or labels mitigate impact of clothes. However, relying unpractical...

Existing person re-identification (Re-ID) methods principally deploy the ImageNet-1K dataset for model initialization, which inevitably results in sub-optimal situations due to large domain gap. One of key challenges is that building large-scale Re-ID datasets time-consuming. Some previous efforts address this problem by collecting images from internet e.g., LUPerson, but it struggles learn unlabeled, uncontrollable, and noisy data. In paper, we present a novel paradigm Diffusion-ReID...

Cloth-changing person Re-IDentification (Re-ID) is a particularly challenging task, suffering from two limitations of inferior discriminative features and limited training samples.Existing methods mainly leverage auxiliary information to facilitate identity-relevant feature learning, including softbiometrics shapes or gaits, additional labels clothing.However, this may be unavailable in real-world applications.In paper, we propose novel FInegrained Representation Recomposition (FIRe 2 )...