A5035236015- 3D Shape Modeling and Analysis
- Generative Adversarial Networks and Image Synthesis
- Face recognition and analysis
- Computer Graphics and Visualization Techniques
- Advanced Vision and Imaging
- Image Enhancement Techniques
- Advanced Numerical Analysis Techniques
- Advanced Image Processing Techniques
- Face and Expression Recognition
- Optical Wireless Communication Technologies
- Image and Object Detection Techniques
- Human Pose and Action Recognition
- Human Motion and Animation
- Medical Image Segmentation Techniques
- Industrial Vision Systems and Defect Detection
- Textile materials and evaluations
ETH Zurich
2022-2024
University of Wisconsin–Madison
2024
Walt Disney (Switzerland)
2024
Hubei University Of Economics
2024
Zhejiang University
2018-2021
2D portrait animation has experienced significant advancements in recent years. Much research utilized the prior knowledge embedded large generative diffusion models to enhance high-quality image manipulation. However, most methods only focus on generating RGB images as output, and co-generation of consistent visual plus 3D output remains largely under-explored. In our work, we propose jointly learn appearance depth simultaneously a diffusion-based generator. Our method embraces end-to-end...
Undoubtedly, high-fidelity 3D hair plays an indispensable role in digital humans. However, existing monocular modeling methods are either tricky to deploy systems (e.g., due their dependence on complex user interactions or large databases) can produce only a coarse geometry. In this paper, we introduce NeuralHDHair, flexible, fully automatic system for from single image. The key enablers of our two carefully designed neural networks: IRHairNet (Im-plicit representation using network)...
We introduce a deep learning based framework for modeling dynamic hairs from monocular videos, which could be captured by commodity video camera or downloaded Internet. The mainly consists of two neural networks, i.e., HairSpatNet inferring 3D spatial features hair geometry 2D image features, and HairTempNet extracting temporal motions frames. are represented as occupancy fields depicting the volume shapes orientation indicating growing directions. bidirectional warping fields, describing...
Active soft bodies can affect their shape through an internal actuation mechanism that induces a deformation. Similar to recent work, this paper utilizes differentiable, quasi-static, and physics-based simulation layer optimize for signals parameterized by neural networks. Our key contribution is general implicit formulation control active defining function enables continuous mapping from spatial point in the material space value. This property allows us capture signal's dominant...
Physically-based simulation is a powerful approach for 3D facial animation as the resulting deformations are governed by physical constraints, allowing to easily resolve self-collisions, respond external forces and perform realistic anatomy edits. Today's methods data-driven, where actuations finite elements inferred from captured skin geometry. Unfortunately, these approaches have not been widely adopted due complexity of initializing material space learning deformation model each character...
3D facial animation is often produced by manipulating deformation models (or rigs), that are traditionally parameterized expression controls. A key component usually overlooked 'style', as in, how a particular performed. Although it common to define semantic basis of expressions characters can perform, most perform each in their own style. To date, style entangled with the expression, and not possible transfer one character another when considering animation. We present new face model, based...
Abstract Controlling stroke size in Fast Style Transfer remains a difficult task. So far, only few attempts have been made towards it, and they still exhibit several deficiencies regarding efficiency, flexibility, diversity. In this paper, we aim to tackle these problems propose recurrent convolutional neural subnetwork, which call stroke‐pyramid , control the Transfer. Compared state‐of‐the‐art methods, our method not achieves competitive results with much fewer parameters but provides more...
In this paper, we present iOrthoPredictor, a novel system to visually predict teeth alignment in photographs. Our takes frontal face image of patient with visible malpositioned along corresponding 3D model as input, and generates facial aligned teeth, simulating real orthodontic treatment effect. The key enabler our method is an effective disentanglement explicit representation the geometry from in-mouth appearance, where accuracy transformation ensured by while appearance modeled latent...
Physically-based simulation is a powerful approach for 3D facial animation as the resulting deformations are governed by physical constraints, allowing to easily resolve self-collisions, respond external forces and perform realistic anatomy edits. Today's methods data-driven, where actuations finite elements inferred from captured skin geometry. Unfortunately, these approaches have not been widely adopted due complexity of initializing material space learning deformation model each character...
We present a quasi-static finite element simulator for human face animation. model the as an actuated soft body, which can be efficiently simulated using Projective Dynamics (PD). adopt Incremental Potential Contact (IPC) to handle self-intersection. However, directly integrating IPC into simulation would impede high efficiency of PD solver, since stiffness matrix in global step is no longer constant and cannot pre-factorized. notice that actual number vertices affected by collision only...