Nupur Thakur

Papers

Total Citations

H-Index

About

Nupur Thakur’s research focuses on the security and reliability of deep neural networks, particularly in computer vision and video analysis. Her most cited work, “Evaluating a Simple Retraining Strategy as a Defense Against Adversarial Attacks” (2020, 4 citations), addresses a critical vulnerability in DNNs: their susceptibility to small, imperceptible perturbations that can cause misclassification. By systematically testing a straightforward retraining approach, Thakur provides practical insights into defending models against such adversarial threats—a key concern for real-world AI deployment. Her more recent work, “Unsupervised Action Anticipation Through Action Cluster Prediction” (2025, 1 citation), tackles the challenging task of predicting near-future human actions in videos without labeled data. This research has implications for human-helping robotics, collaborative AI, and surveillance, as it seeks to decode complex spatiotemporal dynamics. Though early in her career, Thakur’s contributions bridge foundational security concerns with cutting-edge video understanding, demonstrating a commitment to making AI both robust and context-aware. Her work is a valuable resource for students and researchers exploring adversarial robustness and unsupervised temporal reasoning.