Mohammadreza Mofayezi

I am Mohammadreza, a senior student pursuing a B.Sc. in Computer Engineering at Sharif University of Technology.

My research interests lie in the intersection of computer vision and natural language processing. Specifically, I like investigating different aspects of generative modeling and multi-modal learning in these fields. I am also intrigued by the concept of reasoning and robustness.

From 2020 to 2022, I worked as an Augmented Reality Developer at the MadLlama game studio. I was responsible for developing mobile AR games with ARCore, ARKit and Unity.

Afterwards, I worked under the supervision of Prof. Alahi and the VITA lab at EPFL on human motion prediction. We proposed a unified novel generative diffusion model for human motion reconstruction and prediction from incomplete and noisy data. Our research resulted in a paper published in ICRA’23 and NeurIPS 2022 Workshop on Score-Based Methods: A generic diffusion-based approach for 3D human pose prediction in the wild.

Then, under the supervision of Prof. Kortylewski and the GVRL lab at MPI, I conducted research on the robustness of image classifiers. We proposed a novel benchmark for evaluating the robustness of image classifiers to text‑guided corruptions. Our work was accepted to CVPR’23 Workshop on Generative Models for Computer Vision: Benchmarking Robustness to Text-Guided Corruptions.

I am currently working on my B.Sc. thesis under the supervision of Dr. Asgari, focusing on multi-modal multilingual human face generation and editing. I intend to develop a framework for face generation and editing by immediately editing facial landmarks and semantic segmentation. I’m also creating a large-scale multi-modal multilingual face dataset with more than 200K images.