Shagun Uppal - Research

Research

I'm interested in robotics, computer vision, deep learning, and reinforcement learning. Representative papers are highlighted.

Conference Papers
	SPIN: Simultaneous Perception, Interaction and Navigation Shagun Uppal, Ananye Agarwal, Haoyu Xiong, Kenny Shaw, Deepak Pathak, CVPR, 2024 project page / arXiv An end-to-end reactive mobile manipulation framework for robots showcasing whole-body coordination and an active visual system.
	Dextrous Functional Grasping Ananye Agarwal, Shagun Uppal, Kenny Shaw, Deepak Pathak, CoRL, 2023 project page / arXiv A modular appraoch for dextrous functional grasping for in-the-wild objects using a modular approach of passive data and sim2real learning.
	Emotional Talking Faces: Making Videos More Expressive and Realistic Sahil Goyal, Shagun Uppal, Sarthak Bhagat, Yi Yu, Yifang Yin, Rajiv Ratn Shah, ACM MM Asia, 2022; Workshop on AI for Creative Video Editing and Understanding, ICCV 2023 Workshop on Multimedia Content Generation and Evaluation: New Methods and Practice (McGE), MM 2023 Best Demo Paper Award; 300+ stars on GitHub project page / code / video / paper A framework for generating expressive talking face videos with controllable emotions and realistic facial animations.
	Suspect Identification Framework using Contrastive Relevance Feedback Devansh Gupta, Aditya Saini, Sarthak Bhagat, Shagun Uppal, Rishi Raj Jain, Drishti Bhasin, Ponnurangam Kumaraguru, Rajiv Ratn Shah, CoRL, 2023 paper / video A framework for suspect identification using contrastive relevance feedback to improve search accuracy through iterative refinement.
	Distilling Motion Planner Augmented Policies into Visual Control Policies for Robot Manipulation I-Chun Arthur Liu, Shagun Uppal, Gaurav S. Sukhatme, Joseph J. Lim, Peter Englert, Youngwoon Lee, CoRL, 2021 project page / arXiv A policy distillation approach (MoPA-PD) for solving complex manipulation tasks in obstructed environments using visual observations in real-time.
	Disentangling Multiple Features in Video Sequences using Gaussian Processes in Variational Autoencoders Sarthak Bhagat, Shagun Uppal, Zhuyun Yin, Nengli Lim, ECCV, 2020 paper / code / video A variational autoencoder (MGP-VAE) which uses Gaussian processes (GP) to model the latent space for the unsupervised learning of disentangled representations in video sequences.
	PrOSe: Product of Orthogonal Spheres Parameterization for Disentangled Representation Learning Ankita Shukla, Shagun Uppal, Sarthak Bhagat, Saket Anand, Pavan Turaga, BMVC, 2019 paper Learning disentangled latent space representations as a parameterized product of orthogonal spheres for images.
	C3VQG: Category Consistent Cyclic Visual Question Generation Shagun Uppal, Anish Madan, Sarthak Bhagat, Yi Yu, Rajiv Ratn Shah ACM MM Asia, 2020; CVPR Workshop on VQA and Dialogue, 2020 paper / code / video A weakly-supervised VAE setting for question generation using category consistent cyclic loss and structured latent space constraints.
	Geometry of Deep Generative Models for Disentangled Representations Ankita Shukla, Sarthak Bhagat, Shagun Uppal, Saket Anand, Pavan Turaga, ICVGIP, 2018 paper Analysis of latent spaces of different models of disentanglement under the lens of Riemannian Geometry.
Journal Papers
	Multimodal Research in Vision and Language: A Review of Current and Emerging Trends Shagun Uppal, Sarthak Bhagat, Devamanyu Hazarika, Navonil Majumdar, Soujanya Poria, Roger Zimmermann, Amir Zadeh A detailed overview of the latest trends in research pertaining to visual and language modalities.
Workshop Papers
	DisCont: Self-Supervised Visual Attribute Disentanglement using Context Vectors Sarthak Bhagat, Vishaal Udandarao, Shagun Uppal* PTSGM Workshop, ECCV, 2020; MLI4SD Workshop, ICML, 2020 A selfsupervised framework to disentangle multiple factors of variation by exploiting the structural inductive biases within images.