Ankan Dash
PhD - Computer Science
Specializing in AI, Deep Learning, Computer Vision, LLMs and Generative AI. Passionate about developing AI systems that benefit society while advancing the frontiers of computational intelligence.
Specializing in AI, Deep Learning, Computer Vision, LLMs and Generative AI. Passionate about developing AI systems that benefit society while advancing the frontiers of computational intelligence.
New Jersey Institute of Technology
2020 - 2025
Focus: Deep Learning, Computer Vision, Generative AI
University of Illinois at Urbana-Champaign
Finite Element Methods, CFD models for flow field analysis
KIIT University
CFD, Machine Learning, FEM
Samsung Research, Mountain View, CA
Successfully developed real-time reverse pass-through feature for Samsung VR.
Samsung Research, Mountain View, CA
Developed POC for reverse pass-through feature and Diffusion Models for image inpainting.
A showcase of my key research and development projects, highlighting practical applications of my work in computer science and AI.
RevAvatar is an AI-powered framework that enables reverse pass-through in VR headsets by reconstructing high-fidelity 2D facial images and generating accurate 3D head avatars from partially visible facial regions. By leveraging advanced generative models, RevAvatar enhances visual communication and enables more natural, immersive interactions in virtual environments.
GANeRFine is a hybrid framework that enhances NeRF-based 3D scene reconstruction by integrating GANs to improve texture fidelity and fine details, even from sparse inputs. Compatible with multiple NeRF variants, it delivers photorealistic results and near real-time performance, making it ideal for immersive applications such as virtual environments, avatars, and Metaverse content creation.A GAN-based framework designed to enhance the output quality of any NeRF-based model.
During my internship at Samsung in 2023 and 2024, I worked on developing a novel reverse pass-through system for VR headsets, enabling real-time facial reconstruction from occluded views. I also contributed to building fast, real-time face swapping and image colorization models, as well as diffusion-based models for image inpainting and restoration.
Right image: (TOP) Apple's EyeSight vs (BOTTOM) my approach output.
International Joint Conference on Artificial Intelligence (IJCAI) • 2025
This paper presents RevAvatar, a reverse pass-through framework that leverages generative models to reconstruct high-fidelity 2D faces and synthesize 3D head avatars from partially occluded inputs in VR, enabling more natural social interaction and enhancing immersive communication.
Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL) • 2025
This paper presents TEXTFLOW, a modular vision-language system that combines VLMs for extracting structured text from flowchart images and LLMs for reasoning over the text, enhancing controllability, explainability, and overall performance in flowchart understanding.
IEEE International Conference on Metaverse Computing, Networking, and Applications • 2025
This paper presents GANeRFine, a framework that uses GANs to refine NeRF outputs, improving photorealism and detail in 3D scene reconstruction from sparse inputs for real-time, high-fidelity rendering.
IEEE • 2024
This paper provides a comprehensive survey of GAN theory, variants, evaluation metrics, and their diverse applications across STEM, business, and arts, offering researchers a broad overview to apply GANs in various fields.
Download my complete academic CV or view key highlights below.
Comprehensive overview of my education, research experience, publications, awards, and professional activities in computer science and AI.
Recognized at NJIT Graduate Research Day (2025)
“Eye-See-You: Reverse Pass-Through VR and Head Avatars” to appear at IJCAI-2025
“Beyond End-to-End VLMs” accepted for oral presentation at NAACL-2025
AI-ML Computer Vision Intern MPS Group (2024)
“Self-Supervised Learning for User Localization” presented at ICNC 2024
“Hier-GAN: Hierarchical Inpainting GAN with Auxiliary Inputs” accepted to CVPR – Computer Vision in the Wild (2024)
AI-ML Intern (2023)
Playing instruments and curating playlists — music keeps life in rhythm.
A amazing cook — always experimenting with bold flavors and restaurant-level meals at home.
Passionate about exploring new places, cultures, and cuisines — both near and far.
From theory to breakthroughs — following the evolving world of AI with genuine interest.
I'm always interested in discussing research collaborations, speaking opportunities, and potential partnerships.