Anubhav Gupta

Portrait


Hi!

I am a Ph.D. student in Computer Science at the University of Maryland, College Park advised by Prof. Abhinav Shrivastava. I am currently working on understanding videos through graphs and my plan is to apply these techniques to understand long videos. I have published works on understanding style copying in diffusion models; showing semantic capabilities with implicit neural represenataions (details below).

Before starting my graduate studies, I worked in industry for 8 years in various research and engineering roles. I have primarily focused on perception and my most relevant experience in this domain has been in autonomous vehicles space. In my previous life, I worked as an Analyst with multiple banks. I received my bachelors from IIT Delhi.

News

Research Highlights

Style similarity in diffusion models

Preference feedback collected by human or VLM annotators is often noisy, presenting a significant challenge for preference-based reinforcement learning. To address this challenge, we propose TREND, a novel framework that integrates few-shot expert demonstrations with a tri-teaching strategy for effective noise mitigation.



Style similarity in diffusion models

Can we measure the style similarity between images? We propose a way to extract style from images. We call this Contrastive Style Descriptors (CSD). Using this model, we study the style replication in image generation models.



Latent-INR: A Flexible Framework for Implicit Representations of Videos with Discriminative Semantics

We show semantic capabilities in Implicit Neural Representations (INR) by proposing a novel framework that learns discriminative semantics in videos.



LEIA: Latent View-invariant Embeddings for Implicit 3D Articulation

Modeling unseen 3D articulation states by interpolating across a learnable, view-invariant latent embedding space.



PatchGame overview

Emergent communication via mid-level patches in a referential game played on a large-scale image dataset.



Points of Interest overview
Mining points of interest via address embeddings: an unsupervised approach
LocalRec '21: Proceedings of the 5th ACM SIGSPATIAL

Unsupervised PoI mapping (polygon boundaries) using GPS, OpenStreetMaps and Address Information in highly dense environments

Abhinav Ganesan, Anubhav Gupta, Jose Mathew


OW overview
Two Party Evaluation of the Open Visual World
Under Review

Defining a paradigm for unbiased evaluation in open worlds.