معرفی
Shashank Agnihotri is a Researcher & PhD Candidate at the Chair for Machine Learning within the Data and Web Science Group at the University of Mannheim. He is affiliated with the School of Business Informatics and Mathematics. His research focuses on adversarial and OOD robustness of vision models, pixel-wise prediction tasks, and neural architecture search. He has contributed to projects like CosPGD (an efficient adversarial attack for pixel-wise tasks) and Improving Feature Stability during Upsampling.
Education:
- PhD Candidate in Computer Science, University of Mannheim (2023–Present)
- PhD Candidate in Computer Science, University of Siegen (2022–2023)
- MSc. Computer Science, Albert-Ludwigs Universität Freiburg (2018–2021)
- B.E. Computer Engineering, VESIT, University of Mumbai (2014–2018)
Research Interests:
- Adversarial and OOD robustness of deep learning models
- Sensor layout optimization and task-specific camera parameters
- Signal processing impact on model reliability
- Neural architecture search (NAS) methodologies
Publications & Awards:
- Published at ICML, ECCV, ICCV, NeurIPS, and ICCP
- Outstanding Reviewer (CVPR 2025) and Notable Reviewer (ICLR 2025)
- Organized the 45th DAGM German Conference on Pattern Recognition (GCPR 2023)
Labs & Collaborations: Part of the Machine Learning Group at Mannheim and previously at the Machine Learning Group in Freiburg under Prof. Frank Hutter and Prof. Thomas Brox.
۰مقاله منتشرشده
