Shashank Agnihotri

I'm a Ph.D. candidate at the Chair for Machine Learning supervised by Prof. Dr.-Ing Margret Keuper.

My research interests are:

• Architectural Design Choices for OOD Robustness and Adversarial Robustness of Vision Models
• Pixel-wise prediction tasks
• Neural Architecture Search.

My Resume

Shashank Agnihotri is currently a Ph.D. student at Prof. Dr. Ing Margret Keuper's Machine Learning Group at the University of Mannheim. 
Before this, he worked as a Research Assistant in the Machine Learning Group at Albert-Ludwigs Universität Freiburg, with Prof. Dr. Frank Hutter and Dr. Mahmoud Safari, working on Neural Architecture Search (NAS) and Robustness of various one-shot NAS methods. 
He received his MSc. Computer Science from Albert-Ludwigs Universität Freiburg in September 2021 and was supervised by Dr. Tonmoy Saikia and Prof. Dr. Thomas Brox. 
In July 2018, he completed his Bachelors in Engineering (B.E.) Computer Engineering from Vivekanand Education Society's Institute of Technology (VESIT), University of Mumbai.
He works primarily on analyzing and improving the adversarial and OOD robustness of deep learning methods. 
Additionally, he is interested in sensor and sensor-related parameters and signal processing, and its impact on the robustness and reliability of Deep Learning based models. He has published some papers at prestigious venues like ICML, ECCV, ICCV, NeurIPS, and ICCP and served as a reviewer for TPAMI, Nature Scientific Reports, ICCV, ECCV, NeurIPS, ICLR, BMVC, and other top-tier venues. 
He helped in the organization of the 45th DAGM German Conference on Pattern Recognition (GCPR) 2023, Heidelberg. 

• CosPGD: An efficient white-box adversarial attack for pixel-wise prediction tasks – Accepted at ICML 2024
• Improving Feature Stability during Upsampling -- Spectral Artifacts and the Importance of Spatial Context – Accepted at ECCV 2024

• December 2023 – Present : P.h.D. Candidate, Computer Science, University of Mannheim
• June 2022 – December, 2023 : P.h.D. Candidate, Computer Science, University of Siegen
• October 2018 – September 2021 :  MSc. Computer Science, University of Freiburg
• July 2014 – June 2018 : BE Computer Engineering, VESIT, Univeristy of Mumbai
• June 2000 – May 2014 : Smt. Sulochanadevi Singhania School  

• Gavrikov, P., Agnihotri, S., Keuper, M. and Keuper, J., 2024. “How Do Training Methods Influence the Utilization of Vision Models?” accepted at the Interpretable AI: Past, Present and Future Workshop at NeurIPS 2024
• Agnihotri, Shashank, Julia Grabinski, and Margret Keuper. “Improving Feature Stability during Upsampling -- Spectral Artifacts and the Importance of Spatial Context” accepted at the European Conference on Computer Vision (ECCV) 2024.Agnihotri, S., Jung, S. & Keuper, M.. (2024).
• "CosPGD: an efficient white-box adversarial attack for pixel-wise prediction tasks". Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:416–451 Available from https://proceedings.mlr.press/v235/agnihotri24b.html.
• Sommerhoff, Hendrik, Shashank Agnihotri, Mohamed Saleh, Michael Moeller, Margret Keuper, and Andreas Kolb. “Differentiable Sensor Layouts for End-to-End Learning of Task-Specific Camera Parameters.” accepted at the Internation Conference on Computational Photography (ICCP) 2024.
• Agnihotri, Shashank, Kanchana Vaishnavi Gandikota, Julia Grabinski, Paramanand Chandramouli, and Margret Keuper. “On the unreasonable vulnerability of transformers for image restoration-and an easy fix.” In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 3707-3717. 2023.
• Hoffmann, Jasper*, Shashank Agnihotri*, Tonmoy Saikia, and Thomas Brox. “Towards improving robustness of compressed CNNs.” In ICML Workshop on Uncertainty and Robustness in Deep Learning (UDL). 2021.