RecMoDiffuse: Recurrent Flow Diffusion for Human Motion Generation paper put on arXiv
We build a new recurrent diffusion model for human motion generation.
👨🏻💻About Me
Mirgahney H. Mohamed is a PhD student at UCL Foundational AI CDT. His research interests include 3D computer vision, motion and deformation, representation learning, and generative modelling. Mirgahney is an active learner seeking knowledge in everything and from everyone, with a great enthusiasm for AI, machine learning and other topic in science including mathematics, and physics. When he isn’t glued to a laptop screen, he spends time reading books, learning foreign language, and trying to use projectiles in playing football.
Interests
- 3D Computer Vision
- Motion and deformation
- Representation Learning
- Generative Modeling
- Uncertainty Estimation
Education
PhD Artificial Intelligence (Current)
University College London (UCL)
MSc Machine Intelligence
AIMS - African Master in Machine Intelligence
BSc Statistics and Computer Science
University of Khartoum, Faculty of Mathematical Science
📚 My Research
I’m a PhD student in Computer Vision at University College London (UCL). My research interests include 3D computer vision, motion and deformation, representation learning, and generative modeling. I’m particularly interested in advancing generative models in 4D computer vision models.
I work with various deep learning techniques to model the 4D world, including NeRF, Gaussian Splatting and Diffusion models.
Please reach out to collaborate 😃
Featured Publications
Recent Publications
(2024).
Dynamicsurf: Dynamic neural rgb-d surface reconstruction with an optimizable feature grid.
In ICW.
(2024).
RecMoDiffuse: Recurrent Flow Diffusion for Human Motion Generation.
In ICW.
(2022).
GNPM: Geometric-aware neural parametric models.
In ICW.
Recent & Upcoming Talks
Recent News
Start Student Researcher internship at Google DeepMind
Work on 4D vision project with the foundational research team at DeepMind.
Presented DynamicSurf at 3DV 2024 in Davos, Switzerland.
We propose a model-free neural implicit surface reconstruction method for high-fidelity 3D modelling of non-rigid surfaces from monocular RGB-D video.