Arnav Varma

Advanced Engineering (Czechia), TomTom (Netherlands)

Papers

Total Citations

H-Index

About

Arnav Varma is a researcher advancing the frontiers of computer vision, with a focus on the reliability and adaptability of deep learning models for spatial understanding. His work primarily targets monocular depth estimation and pose estimation—critical tasks for autonomous driving and robotics. Varma’s major contributions lie at the intersection of adversarial robustness and continual learning. In his highly cited 2022 paper, "Adversarial Attacks on Monocular Pose Estimation" (8 citations), he exposed vulnerabilities in deep neural networks, highlighting the challenge of deploying these models in safety-critical environments. He further innovated with "Continual Learning of Unsupervised Monocular Depth from Videos" (2024, 7 citations), enabling models to adaptively learn from diverse, crowd-sourced video streams without forgetting prior knowledge—a breakthrough for real-world scalability. His 2023 work, "Transformers in Unsupervised Structure-from-Motion" (2 citations), explores transformer architectures to enhance 3D scene reconstruction. With a growing citation impact and a focus on practical, robust vision systems, Varma’s research is shaping how autonomous agents perceive and navigate dynamic environments, making him a notable voice in modern computer vision.