2024-10: We introduce DreamCraft3D++, a new technique for high-quality 3D content generation!
2024-08: One paper is accepted by SIGGRAPH ASIA 2024!
2024-03: One paper is accepted by SIGGRAPH 2024!
2024-02: Two papers are accepted by CVPR 2024!
2024-02: Started my internship at NVIDIA Research with Koki Nagano and Shalini De Mello.
2024-01: DreamCraft3D is accepted at ICLR 2024. See you in Vienna!
2023-09: HAvatar is accepted by ACM TOG 2023.
2023-03: StyleAvatar is accepted by ACM SIGGRAPH 2023.
2023-03: Next3D is selected as a CVPR Highlight paper (top 10% of accepted papers).

Introducing DeepSeek-VL, an open-source Vision-Language (VL) Model designed for real-world vision and language understanding applications.

We present DreamCraft3D++, an extension of DreamCraft3D that enables efficient, high-quality generation of complex 3D assets in just 10 minutes.

Given a reference image, SMPL sequences, and camera parameters, our method can generate free-view dynamic human videos in 360 degrees.

We present DreamCraft3D, a hierarchical 3D content generation method that produces high-fidelity and coherent 3D objects.

We present the Incremental 3D GAN Inversion approach, which reconstructs photorealistic 3D facial avatars in under one second from single or multiple source images.

We propose Control4D, which enables high-fidelity and spatiotemporally consistent 4D portrait editing from only text instructions.

We present RAM-Avatar, a real-time photorealistic human avatar method from monocular videos that supports high-fidelity rendering, full-body control (face and hands), and real-time animation.

We introduce VectorTalker, a novel method for creating high-fidelity, audio-driven talking heads using scalable vector graphics, effective for various styles.

We introduce the Facial Model Conditioned Neural Radiance Field, a hybrid 3D representation that combines the flexibility of NeRF with a parametric template, leveraging synthetic renderings for conditioning.

We propose StyleAvatar, a real-time photorealistic portrait avatar reconstruction approach using StyleGAN-based networks, yielding high-fidelity avatars with faithful expression control.

We propose a novel 3D GAN framework for unsupervised learning of high-quality, 3D-consistent facial avatars from unstructured 2D images. Our approach introduces Generative Texture-Rasterized Tri-planes for accurate deformations and topological flexibility.

We propose an efficient method to build a personalized generative prior from a small set of facial images of a specific individual. This approach supports photo-realistic novel view synthesis and face reenactment using various input signals (images, 3DMM coefficients, or audio).

We present DiffuStereo, a novel system using only sparse cameras (8 in our setting) for high-quality 3D human reconstruction. Central to our method is a diffusion-based stereo module, which incorporates powerful generative diffusion models into iterative stereo matching.

We introduce a high-resolution, 3D-aware generative model that supports local control over facial shape and texture, with real-time interactive editing capabilities.

We propose FENeRF, a 3D-aware generator that produces view-consistent, locally-editable portraits. By employing two decoupled latent codes for facial semantics and texture, we achieve flexible and precise manipulation of geometry and appearance.

We propose a novel optimization-based framework for multi-view motion capture from internet videos, recovering more precise and detailed poses than monocular pose estimation methods.

We present a general and practical evaluation framework for multiple widely used bus-arrival-time prediction models, including delay-based, k-NN, kernel regression, additive models, and LSTM-based neural networks.

This website template was adapted from Yu Deng.