To address the efficiency bottlenecks in hierarchical GNN-based cloth simulation, we propose two targeted architectural components that jointly optimize message propagation and topological redundancy. Specifically, we introduce a Redundant Message Reduction strategy that streamlines cloth-body interactions by restricting communication to only necessary stages, and a Proportional Edge Pruning mechanism that adaptively reduces the number of incoming edges per node while retaining sufficient deformation paths. These designs substantially reduce computational overhead with minimal sacrifice of physical fidelity. Extensive experiments across diverse garments and motion sequences demonstrate that our model achieves more than a 40% increase in inference speed and a 22% reduction in training time, while ensuring both numerical and visual fidelity. In addition, qualitative evaluations confirm that our method preserves realistic garment dynamics under challenging poses and unseen clothing structures.

High-fidelity garment reconstruction is essential for various applications such as garment design and virtual try-on. While image-based reconstruction methods have made significant progress with deep generative models, generating 3D models from hand-drawn sketches to meet design intentions remains challenging. One of the main obstacles is the limited availability of large-scale 3D garment models accompanied by corresponding sketches. To address this issue, we propose SGM, a comprehensive dataset comprising 656 garment models categorized into short and long sleeves. Each garment model in SGM is accompanied by four types of rendered images and a series of UDF values. Furthermore, we introduce a novel baseline approach for sketch-based garment reconstruction using an end-to-end generative network capable of generating garment models from single hand-drawn sketches. Extensive experimental results highlight the significance and value of our proposed dataset and method. We plan to make SGM publicly available upon publication.

3D human reconstruction from a single image has achieved great progress with recent deep neural networks. However, conventional approaches still struggle with the issues of over-smoothing details and wrong limb poses. To this end, we propose PMDI, a method that combines parametric-model and depth-aware implicit function for single-view human reconstruction. First, given an RGB image of a person’s whole body as input, the method predicts its corresponding SMPL parameter model, depth map, and front (back) normal map by using deep neural networks. Then, the predicted front depth map and normal feature are used as the additional parameters of the deep implicit function for reconstructing coarse results. Finally, the fine result is produced by integrating its corresponding coarse result with detailed back D-BiNI surface. Extensive experiments on the current large publicly available dataset (including DeepHuman and THUman2.0) demonstrate that PMDI outperforms the state-of-the-arts including PIFu, PIFuHD,PaMIR, and ICON.

Abstract Mesh denoising is a fundamental problem in geometry processing. The main challenge is to preserve sharp features (such as edges and corners) and smooth regions (such as smoothly curved regions and fine details) while removing the noise. State‐of‐the‐art denoising methods still struggle with this issue. In this paper, we first propose a new variational model combining total variation and anisotropic Laplacian regularization to filter the normal vector field of the mesh. This model can preserve sharp features and simultaneously handle smooth regions well. Then, a new vertex updating scheme is presented to reconstruct the mesh according to the filtered face normals. It prevents the orientation ambiguity problem introduced by existing schemes. Experiments show that our denoising method outperforms all compared methods visually and quantitatively, especially for meshes consisting of both sharp features and smooth regions.