Tarun Yenamandra

PhD candidate at TU Munich (advised by Prof. Daniel Cremers), specializing in the intersection of Generative AI and 3D/4D Computer Vision. My research bridges 3D geometry with conditional generation — from implicit morphable models and single-pass SDF rendering to learning-free sparse-view reconstruction. Previously Research Scientist Intern at Meta Reality Labs Research (diffusion-based scene editing, dynamic 3DGS).

tarun738 at gmail dot com  /  Scholar  /  LinkedIn  /  GitHub

News

• Jun 2026 — VisDom preprint on arXiv
• Jul–Dec 2025 — Research Scientist Intern at Meta Reality Labs Research (work in submission)

Selected Papers

My interests are at the intersection of generative AI and 3D/4D computer vision — specifically conditional generation with multi-view consistency, neural scene representations (NeRF, 3DGS), and sparse-view reconstruction. Full list on Google Scholar.

A learning-free, plug-and-play visible domain constraint for sparse novel view synthesis that eliminates geometric ambiguity from limited viewpoints.

Directional and signed-distance function based renderer that enables fast inverse rendering.

Template-free implicit model for full human heads including hairstyle using signed distance fields.

Modeling inverse kinematics as a semidefinite program to relax non-convex terms and remove initialization dependence.