Ming He

About Me

I'm a Computer Science Ph.D. candidate at the University of Sheffield's Department of Computer Science, working under Dr Steve Maddock and Dr Zhixiang Chen's supervision.

Research Interests

• 3D Reconstruction: Reconstruct 3D objects with multi-view images / single-view image
• Controllable Text-to-3D Generation: Lifting text-to-2D generation to 3D domain with multiple conditions

News

• [2025.5] I am serving as Technical Program Chair of BMVC 2025 in Sheffield, UK.

Ongoing Research

• SIC3D: Controllable Text- and Image-Conditioned 3D Object Generation SIC3D is a controllable 3D generation framework based on 3D Gaussian Splatting, designed to improve fidelity, style consistency, and user control in text-to-3D synthesis. It adopts a two-stage pipeline: a Geometry Generation Stage using Variational Score Distillation (VSD) to produce base geometry from text, followed by a Style Distillation Stage that refines the appearance using a novel Variational Stylized Score Distillation (VSSD) loss guided by a 2D reference image. This staged design enables better disentanglement of structure and style, ensuring accurate semantic geometry and visually aligned textures. SIC3D sets a new benchmark in stylized 3D generation and offers a scalable solution for creative content creation.

the Ancient Egyptian Pyramid

A handbag

A furry teddy bear

Publications and Projects