Anirud Nandakumar

I'm a Ph.D. student in the Department of Electrical and Computer Engineering (ECE) at the National University of Singapore (NUS), where I will be advised by Prof. Rajat Talak, working on computer vision for robotics.

I completed B.Tech in Civil Engineering + M.Tech in Data Science (Interdisciplinary dual degree, 5 year programme) at Indian Institute of Technology, Madras (IIT Madras), where I worked with Prof. Kaushik Mitra in the Computational Imaging Lab and with Prof. Lelitha Devi Vanajakshi on intelligent transportation systems.

I also worked as a Visiting Student Researcher at the University of California, Riverside, where I worked with Prof. Vishwanath Saragadam and Prof. Amit K. Roy-Chowdhury on 3D world simulation for autonomous vehicles using RGB, thermal, and LiDAR data.

My broad research interests lie in physics grounded computer vision for Robotics, computational imaging, 3D reconstruction, and reinforcement learning, with applications to robotics and autonomous systems.

Email  /  CV  /  Scholar  /  Github  /  ORCID  /  LinkedIn

My research lies at the intersection of computer vision, computational imaging, and robotics, with a focus on 3D reconstruction, novel view synthesis, and reinforcement learning for autonomous systems. I'm broadly interested in building systems that can perceive and act robustly in challenging real-world environments — from low-light and thermal imaging to large-scale traffic and driving scenarios.

*Equal contribution. Introduces a neuromorphic (event camera) data-guided pipeline for deblurring and HDR novel view synthesis.

Introduces a novel RL-based traffic signal control method to minimize queue lengths, integrated with the SUMO traffic simulator.

A new RGBT YOLOv8 architecture for traffic object detection, comparing pixel-level, feature-level, and decision-level fusion strategies.

A patent on intelligent traffic analysis combining aerial and ground-based RGB and thermal camera feeds.

In preparation for submission to a top venue.

In preparation for submission to a top venue.

Introduced an adversarial reverse-imagination-based offline RL framework trained with Soft Actor-Critic, designing reverse dynamics and reverse rollout models to enhance the synthetic dataset for training.

Developed a multi-modal pipeline using frame sampling (optical flow, embeddings) and cascaded LLaVA to generate progressive captions, structured summaries, and detailed narratives directly from video content.

Placed top 4 at IIT Madras. Performed image restoration from noisy images using a UNet-based generator in a cGAN to preserve defect patterns in electronic chips given a defect mask.

Visiting Student Researcher, University of California, Riverside, CA, US
Guided by Prof. Vishwanath Saragadam and Prof. Amit K. Roy-Chowdhury — June 2025 - Present

• Developing a 3D world simulator using RGB, thermal, and LiDAR data for training autonomous vehicles with reinforcement learning.
• Developing algorithms for SfM, 3D Gaussian Splatting, meshing, and novel view synthesis for thermal imaging.

Visiting Student Intern, Purdue University, IN, US
Guided by Prof. Darcy Bullock, Joint Transportation Research Program — May 2024 - July 2024

• Utilized big data analytics and statistical modeling to estimate hard braking patterns via deceleration thresholds.
• Identified anomalies like work zones and access points to enhance traffic safety using normalized braking rates.

Research Intern, Lab for Imaging Sciences and Algorithms, IISc Bangalore, India
Collaboration with Continental Automotive, guided by Prof. Kunal Narayan Chaudhury — Nov 2023 - Feb 2024

• Developed a selective region-based algorithm to enhance visibility in low-light images of outdoor road conditions.
• Designed brightness-based masking to enhance dark regions while preserving bright ones for autonomous driving.

Prof. Vishwanath Saragadam — UC Riverside; advising my work on 3D world simulation using RGB, thermal, and LiDAR data for autonomous driving.
Prof. Amit K. Roy-Chowdhury — UC Riverside; advising my work on 3D world simulation using RGB, thermal, and LiDAR data for autonomous driving.
Prof. Kaushik Mitra — IIT Madras; advising my work on event-camera-guided deblurring and HDR novel view synthesis and 3D Thermal imaging.
Prof. Kunal Narayan Chaudhury — IISc Bangalore; advised my internship on low-light image enhancement for autonomous driving.
Prof. Lelitha Devi Vanajakshi — IIT Madras; advised my research on RL-based traffic signal control and RGB-thermal fusion for traffic systems.

Template adapted from here.