|
Mayur Deshmukh Saarbrucken, Germany I am a Research Assistant at Max Planck Institute for Informatics in Saarbrucken, Germany, and a Master's student in Computer Science at Saarland University. My research focuses on Computer Vision and Deep Learning, Beyond work, I enjoy hiking, photography, and gaming. |
|
News
|
ResearchMy research focus lies at the intersection of Computer Vision and Deep Learning. In particular, I am interested in the following topics:
|
Publications |
|
E-3DPSM: A State Machine for Event-based Egocentric 3D Human Pose Estimation
We introduce E3DPSM (Event-based 3D Pose State Machine), an event-driven continuous pose state machine for egocentric 3D human pose estimation. By aligning motion dynamics with asynchronous event streams, it produces stable, drift-free pose estimates in real time (80 Hz). Our method sets a new state of the art, improving accuracy by up to 19% and temporal stability by 2.7x. |
|
|
GRMM: Real-Time High-Fidelity Gaussian Morphable Head Model with Learned Residuals
Our method, GRMM (Gaussian Residual Morphable Model), generates high-fidelity 3D head avatars in real time by combining a mesh-based morphable prior with learned Gaussian residuals. This design captures fine geometric and appearance details while maintaining disentangled control over identity and expression, enabling photorealistic and expressive facial synthesis. |
ProjectsPhotonPulsePhotonPulse is a C++ physically based renderer with support for cameras, BSDFs, samplers, lights, integrators, and post-processing features such as denoising and bloom. GitHubSceneChatSceneChat is a text-driven scene generation pipeline that turns a natural language description into a structured room layout, previews the layout in Blender, and can generate 3D assets for the objects in the scene. GitHubSemantic Perceptual Image CompressionThis project implements a semantic, perceptual image compression pipeline that uses a convolutional model to identify regions of interest in an image and allocate compression quality accordingly. GitHub |
Education & Experience
|
| © Mayur Deshmukh 2026 / Design: OpenAI Codex |