Publications

EgoFlow: Gradient-Guided Flow Matching for Egocentric 6DoF Object Motion Generation

CVPR, 2026

A flow-matching framework with gradient-guided inference that generates physically plausible 6DoF object trajectories from egocentric videos, reducing collision rates by up to 79%.

Project Page Paper Code Citation

GMT: Goal-Conditioned Multimodal Transformer for 6-DOF Object Trajectory Synthesis in 3D Scenes

3DV, 2026
Huajian Zeng*, Abhishek Saroha*, Daniel Cremers, Xi Wang

A multimodal transformer framework that generates realistic and goal-directed 6-DOF object trajectories by jointly leveraging 3D bounding box geometry, point cloud context, semantic object categories, and target end poses.

Project Page Paper Video Code Citation

Nonisotropic Gaussian Diffusion for Realistic 3D Human Motion Prediction

CVPR, 2025

A latent diffusion model with nonisotropic Gaussian noise aligned to the human skeleton's kinematic structure for realistic 3D motion prediction.

Project Page Paper Code Citation

GAS-NeRF: Geometry-Aware Stylization of Dynamic Radiance Fields

arXiv, 2025

Joint appearance and geometry stylization in dynamic radiance fields, transferring both color and structural details from style images.

Paper Citation

ZDySS: Zero-Shot Dynamic Scene Stylization using Gaussian Splatting

arXiv, 2025

A zero-shot stylization framework for dynamic scenes using Gaussian Splatting that generalizes to unseen style images at inference.

Paper Citation

DiffCD: A Symmetric Differentiable Chamfer Distance for Neural Implicit Surface Fitting

ECCV, 2024

A novel symmetric and differentiable Chamfer distance loss for neural implicit surface fitting that eliminates spurious surfaces without additional regularization.

Paper Code Citation

Enhancing Surface Neural Implicits with Curvature-Guided Sampling and Uncertainty-Augmented Representations

GCPR, 2024
Lu Sang, Abhishek Saroha, Maolin Gao, Daniel Cremers

Curvature-guided sampling and uncertainty-augmented representations for improved neural implicit surface reconstruction.

Paper Citation

Gaussian Splatting in Style

GCPR, 2024

We are the first to employ Gaussian Splatting to solve the task of scene stylization, extending the work of neural style transfer to three spatial dimensions.

Project Page Paper Citation

ResolvNet: A Graph Convolutional Network with multi-scale Consistency

NeurIPS Workshop (New Frontiers in Graph Learning), 2023

A spectral graph convolutional architecture with multi-scale consistency for weighted graphs.

Paper Citation