Virtual try-on is a promising computer vision topic with high commercial value wherein new garment visually worn on person photo-realistic effect. Previous studies conduct their shape and content inference at one stage, employing single-scale warping mechanism relatively unsophisticated mechanism. These approaches have led to suboptimal results in terms of skin reservation under challenging scenarios. To address these limitations, we propose novel virtual method via progressive paradigm...

As bird's-eye-view (BEV) semantic segmentation is simple-to-visualize and easy-to-handle, it has been applied in autonomous driving to provide the surrounding information downstream tasks. Inferring BEV conditioned on multi-camera-view images a popular scheme community as cheap devices real-time processing. The recent work implemented this task by learning content position relationship via Vision Transformer (ViT). However, its quadratic complexity confines only latent layer, leaving scale...

Purpose Varied shapes and sizes of different products with irregular rough surface fragile properties give a challenge to traditional contact gripping. Single Bernoulli grippers are not suited handle objects as the impact center negative pressure force could result in large deformation stress which damage materials, they also have some limitations for gripping small shapes. Thus, this paper aims design non-contact gripper soft, rough-surfaced multi heads, optimal structures parameters....

Depth estimation provides an alternative approach for perceiving 3D information in autonomous driving. Monocular depth estimation, whether with single-frame or multi-frame inputs, has achieved significant success by learning various types of cues and specializing either static dynamic scenes. Recently, these fusion becomes attractive topic, aiming to enable the combined perform well both However, adaptive cue relies on attention mechanisms, where quadratic complexity limits granularity...

In online clothing sales, static model images only describe specific statuses towards consumers. Without increasing shooting costs, it is a subject to display dynamically by synthesizing continuous image sequence between images. This paper proposes novel human synthesis method pose-shape-content inference. the condition of two reference poses, pose interpolated in manifold controlled linear parameter. The transferred into end shape AdaIN and attention mechanism infer target shape. Then...

This paper proposes a novel garment transfer method supervised with knowledge distillation from virtual try-on. Our first reasons the parsing to provide shape prior downstream tasks. We employ multi-phase teaching strategy supervise training of reasoning model, learning response and feature try-on model. To correct error, it transfers back its owner absorb hard in self-study phase. Guided by parsing, we adjust position transferred via STN prevent distortion. Afterward, estimate progressive...

3D anomaly detection plays a crucial role in monitoring parts for localized inherent defects precision manufacturing. Embedding-based and reconstruction-based approaches are among the most popular successful methods. However, there two major challenges to practical application of current approaches: 1) embedded models suffer prohibitive computational storage due memory bank structure; 2) reconstructive based on MAE mechanism fail detect anomalies unmasked regions. In this paper, we propose...

Collaborative filtering is a kind of widely used and efficient technique in various online environments, which generates recommendations based on the rating information his/her similar-preference neighbors. However, existing collaborative methods have some inadequacies revealing dynamic user preference change evaluating recommendation effectiveness. The sparsity input data may further exacerbate this issue. Thus, paper proposes novel neighbor selection scheme constructed context attenuation...

Abstract In the traditional customized product (CP) configuration design system, rules in database usually rely on manual input and maintenance of experienced designers. complex design, are numerous, complex, difficult to understand. The way human based experience consumes plenty resources, which strictly limited by This overreliance those designers has seriously restricted development enterprises. To address this problem, a least recently used dynamic decision tree (LRU-DDT) algorithm for...

The clothing personalized online customization is required to design around the user shape parameters, and traditional anthropometry method obtains parameter with a low speed high error. To tackle this issue, rapid 3D human body reconstruction proposed based on multi-perspective silhouettes for customization. multilevel dilated convolution semantic network (MDS-Net) leveraged extract global local features in implement segmentation. torso extraction (TPE-Net) pose parameters of segmentation...

As bird's-eye-view (BEV) semantic segmentation is simple-to-visualize and easy-to-handle, it has been applied in autonomous driving to provide the surrounding information downstream tasks. Inferring BEV conditioned on multi-camera-view images a popular scheme community as cheap devices real-time processing. The recent work implemented this task by learning content position relationship via vision Transformer (ViT). However, quadratic complexity of ViT confines only latent layer, leaving...

Virtual try-on is a promising computer vision topic with high commercial value wherein new garment visually worn on person photo-realistic effect. Previous studies conduct their shape and content inference at one stage, employing single-scale warping mechanism relatively unsophisticated mechanism. These approaches have led to suboptimal results in terms of skin reservation under challenging scenarios. To address these limitations, we propose novel virtual method via progressive paradigm...