A Generalizable Light Transport 3D Embedding for Global Illumination
SIGGRAPH 2026
A framework to learn a generalizable 3D light transport embedding that approximates global illumination from 3D scene configurations.
Mukund Varma T
I'm a first second third fourth year CSE Ph.D. student at UC San Diego, advised by Professor Ravi Ramamoorthi (and previously also by Hao Su).
I received my bachelor's degree from IIT Madras.
I was fortunate to have spent several summers at Adobe, Nvidia, Apple, and Google with amazing mentors and colleagues. I have also had the opportunity to collaborate with folks from the VITA Group at UT Austin. During my undergrad, I was also associated with the Advanced Geometric Computing Lab, and the Computational Imaging Lab.
I was lucky to be a recipient of the Jacobs School of Engineering Fellowship.
Research
My current research interests lie at the intersection of computer vision, computer graphics and machine learning, specifically to facilitate high quality 3D reconstructions and semantic understanding from multiple view points.
* indicates equal contribution.
PhotonSplat: 3D Scene Reconstruction and Colorization from SPAD Sensors
ICCP 2025
A framework to reconstruct 3D scenes from SPAD binary images, supporting colorization using a single reference blurry image or generative priors.
SkipWriter: LLM-Powered Abbreviated Writing on Tablets
UIST 2024
A framework that converts handwritten strokes of a variable-length prefix-based abbreviation (e.g. "ho a y" as handwritten strokes) into the intended full phrase (e.g., "how are you" in the digital format) based on the preceding context.
GAURA: Generalizable Approach for Unified Restoration and Rendering of Arbitrary Views
ECCV 2024
A generalizable framework for novel view synthesis using degraded input captures containing any imperfection type.
A Construct-Optimize Approach to Sparse View Synthesis without Camera Pose
SIGGRAPH 2024
A camera-free novel view synthesis technique from sparse input views (as few as 3 images of large-scale scenes).
Lift3D: Zero Shot Lifting of Any 2D Vision Model to 3D
CVPR 2024
General framework to lift any pretrained 2D vision model to generate 3D consistent outputs with no additional optimization.
U2NeRF: Unifying Unsupervised Underwater Image Restoration and Neural Radiance Fields
ICLR Tiny Papers 2024
An unsupervised learning pipeline for generalizable novel view synthesis and restoration of underwater scenes by disentangling into individual image formation components.
Enhancing NeRF akin to Enhancing LLMs: Generalizable NeRF Transformer with Mixture-of-View-Experts
ICCV 2023
We scale up generalizable NeRF training by borrowing the concept of mixture of experts from language models.
DeepSIFT: Rethinking Domain Generalization using Invariant Representations
under review
We present empirical evidence that convolutional networks trained on SIFT features improve robustness to unseen out-of-domain data with minimal to no loss in in-domain performance.
Is Attention All That NeRF Needs?
ICLR 2023
We propose a generalizable neural scene representation and rendering pipeline that achieves superior quality compared to previous methods.
Sparse Lottery Tickets are Data Efficient Image Recognizers
NeurIPS 2022 (Spotlight)
Sparse networks identified using iterative magnitude pruning showcase improved data-efficiency and robustness compared to their dense counterparts.
NL Augmenter: A Collaborative Effort to Transform and Filter Text Datasets
NEJLT 2023 (GEM Workshop, IJCNLP 2021)
Collaborative repository of natural language transformations.
BIG Bench: Beyond the Imitation Game Benchmark
TMLR 2023 (WELM Workshop, ICLR 2021)
Collaborative benchmark for measuring and extrapolating the capabilities of language models.
ShapeFormer: A Transformer for Point Cloud Completion
under review
A specialized pipeline for point cloud shape completion that can generalize to synthetic and real partial scans from seen and unseen categorical types.
[Re] On the Relationship between Self-Attention and Convolutional Layers
ReScience-C (MLRC, NeurIPS 2020)
We introduce Hierarchical Attention, a recurrent transformer module that imitates convolution-like operation with significantly lower computational budget.
Point Transformer for Shape Classification and Retrieval of Urban Roof Point Clouds
IEEE-GRSL 2021
We propose a transformer architecture for sparse set learning, e.g. point cloud understanding.
Academic Service
I regularly serve as a reviewer for major computer vision and machine learning conferences, including ICCV, ECCV, CVPR, ICLR, and NeurIPS. I also enjoyed teaching during my undergraduate studies at IIT Madras, where I served as a teaching assistant for the following courses:
- CS6700 Reinforcement Learning, Spring 2023.
- EE5180 Introduction to Machine Learning, Spring 2023.
- EE5179 Deep Learning for Imaging, Fall 2022.
Misc
I enjoy hiking ⛰️ (cuz duh I am a boring computer science kid), used to sketch ✏️ a bit, and love playing most sports, particularly badminton 🏸, tennis 🎾 and soccer ⚽.
