Hemang Chawla

Advanced Engineering (Czechia), Eindhoven University of Technology

Papers

Total Citations

H-Index

About

Hemang Chawla is a researcher advancing the frontiers of 3D scene understanding and robust computer vision, with a primary focus on monocular depth estimation and structure-from-motion. His work addresses critical vulnerabilities in deep learning systems, particularly the susceptibility of pose estimation networks to adversarial attacks—a contribution that has garnered 8 citations and highlights the challenges of reliable deployment in safety-critical applications like autonomous driving. Chawla has made significant strides in unsupervised and self-supervised learning paradigms, pioneering methods for continual learning of monocular depth from videos (7 citations) and developing image masking techniques that enhance robustness in self-supervised depth estimation (5 citations). His exploration of Transformers in unsupervised structure-from-motion (2 citations) represents a novel integration of attention mechanisms into geometric computer vision. Collectively, Chawla’s research tackles the dual challenges of accuracy and resilience in spatial perception, pushing toward models that can learn from diverse, unlabeled video data while defending against adversarial perturbations. His work is particularly impactful for robotics and autonomous systems, where reliable depth perception is paramount.