Generative Adversarial Networks (GANs) dominate the research field in image-based virtual try-on, but have not resolved problems such as unnatural deformation of garments and the blurry generation quality. While the generative quality of diffusion models is impressive, achieving controllability poses a significant challenge when applying it to virtual try-on and multiple denoising iterations limit its potential for real-time applications. In this paper, we propose Controllable Accelerated virtual Try-on with Diffusion Model (CAT-DM). To enhance the controllability, a basic diffusion-based virtual try-on network is designed, which utilizes ControlNet to introduce additional control conditions and improves the feature extraction of garment images. In terms of acceleration, CAT-DM initiates a reverse denoising process with an implicit distribution generated by a pre-trained GAN-based model. Compared with previous try-on methods based on diffusion models, CAT-DM not only retains the pattern and texture details of the inshop garment but also reduces the sampling steps without compromising generation quality. Extensive experiments demonstrate the superiority of CAT- DM against both GAN-based and diffusion-based methods in producing more realistic images and accurately reproducing garment patterns.

Semantic human body reshaping builds a 3D body according to several anthropometric measurements, playing important roles in virtual fitting and human body design. We propose a novel part-based semantic body model for 3D body reshaping. We adopt 20 types of measurements in regard of length and girth information of body shape. Our approach takes any number (1–20) of measurements as input, and generates a 3D human body. Firstly, all missing measurements are estimated from known measurements using a correlation-based method. Then, based on our proposed semantic model, we learn corresponding semantic body parameters which determine a 3D body from measurements. Our model is trained using a database of 4000 registered body meshes which are fitted with scans of real human bodies. Through experiments, we compare our approach with previous methods and show the advantages of our model.

Curve design on surface meshes has wide applications in computer graphics and computer aided design. The key challenge of the problem is to efficiently and robustly handle the manifold constraint which forces the curve to exactly lie on the surface meshes. Popular approaches, such as projection-based and smoothing-based methods, solve it by either totally relaxing or strictly holding the above constraint, which have their own merits. We propose a shell space method which combines the advantages of both approaches. First, a shell space surrounding the mesh with a distance-like scalar field is constructed. Then the manifold constraint is relaxed to the shell space and a global optimization with the interior point method is conducted. Mimicking the behavior of the both methods, it gradually shrinks the shell space and increases the weight of the manifold energy during the iterations. The adaptive scheme sufficiently relaxes the curve to gain its geometric property (e.g., smoothness) easily (like projection-based method) but converges stably and robustly (like smoothing-based method). Finally, the curve is snapped to the surface with a robust projection. Experiments exhibit that our method outperforms existing work on various aspects, including efficiency, robustness, and controllability.

We present an incremental collision handling algorithm for GPU-based interactive cloth simulation. Our approach exploits the spatial and temporal coherence between successive iterations of an optimization-based solver for collision response computation. We present an incremental continuous collision detection algorithm that keeps track of deforming vertices and combine it with spatial hashing. We use a non-linear GPU-based impact zone solver to resolve the penetrations. We combine our collision handling algorithm with implicit integration to use large time steps. Our overall algorithm, I-Cloth, can simulate complex cloth deformation with a few hundred thousand vertices at 2 - 8 frames per second on a commodity GPU. We highlight its performance on different benchmarks and observe up to 7 - 10X speedup over prior algorithms.

We present an accurate and robust algorithm for self-collision detection in deformable models. Our method is based on the normal cone test and is suitable for both discrete and continuous collision queries on triangular meshes. We propose a novel means of employing surface normal cones and binormal cones to perform the normal cone test. Moreover, we combine our culling criteria with bounding volume hierarchies (BVHs) and present a hierarchical traversal scheme. Unlike the previous BVH-based dual-cone method, our method can reliably detect all self-collisions, and it achieves appreciable speedup over other high-level culling methods. (C) 2018 Elsevier Ltd. All rights reserved.

We propose an effective and efficient way to automatically predict clothes size for users to buy clothes online. We take human height and dressed-human silhouettes in front and side views as input, and estimate 3D body sizes with a data-driven method. We adopt 20 body sizes which are closely related to clothes size, and use such 3D body sizes to get clothes size by searching corresponding size chart. Previous image-based methods need to calibrate camera to estimate 3D information from 2D images, because the same person has different appearances of silhouettes (e.g. size and shape) when the camera configuration (intrinsic and extrinsic parameters) is different. Our method avoids camera calibration, which is much more convenient. We set up our virtual camera and train the relationship between human height and silhouette size under this camera configuration. After estimating silhouette size, we regress the positions of 2D body landmarks. We define 2D body sizes as the distances between corresponding 2D body landmarks. Finally, we learn the relationship between 2D body sizes and 3D body sizes. The training samples for each regression process come from a database of 3D naked and dressed bodies created by previous work. We evaluate the whole procedure and each process of our framework. We also compare the performance with several regression models. The total time-consumption for clothes size prediction is less than 0.1 s and the average estimation error of body sizes is 0.824 cm, which can satisfy the tolerance for customers to shop clothes online.

(Bi)harmonic field has wide applications in geometry processing. Traditionally, to locally control the influence region of a (bi)harmonic field, users usually need to determine the range of its support, regions with non-zero scalar values, by prescribing appropriate boundary conditions. However, this way is non-intuitive and inconvenient. We proposed localized quasi-(bi)harmonic field, which is achieved through a ℓ1-norm regularized convex optimization. It can conveniently control the local support of the scalar field while still keeping some nice properties of the (bi)harmonic field. We applied the localized quasi-(bi)harmonic field in applications such as shape deformation and shape merging, and the experiment results show its benefits.

We present an efficient and accurate algorithm for self-collision detection in deformable models. Our approach can perform discrete and continuous collision queries on triangulated meshes. We present a simple and linear time algorithm to perform the normal cone test using the unprojected 3D vertices, which reduces to a sequence point-plane classification tests. Moreover, we present a hierarchical traversal scheme that can significantly reduce the number of normal cone tests and the memory overhead using front-based normal cone culling. The overall algorithm can reliably detect all (self) collisions in models composed of hundreds of thousands of triangles. We observe considerable performance improvement over prior continuous collision detection algorithms.

We present an efficient and robust method which performs well for both strain limiting and treatment of simultaneous collisions. Our method formulates strain constraints and collision constraints as a serial of linear matrix inequalities (LMIs) and linear polynomial inequalities (LPIs), and solves an optimization problem with standard convex semidefinite programming solvers. When performing strain limiting, our method acts on strain tensors to constrain the singular values of the deformation gradient matrix in a specified interval. Our method can be applied to both triangular surface meshes and tetrahedral volume meshes. Compared with prior strain limiting methods, our method converges much faster and guarantees triangle flipping does not occur when applied to a triangular mesh. When performing treatment of simultaneous collisions, our method eliminates all detected collisions during each iteration, leading to higher efficiency and faster convergence than prior collision treatment methods.