Hello!
I'm Tonmoy Saikia, a computer vision and machine learning researcher at Torc Robotics. I completed my PhD in computer vision under the guidance of Prof. Thomas Brox, focusing on methods to improve the robustness and efficiency of deep learning-based computer vision models. To this end, my research explores strategies including data augmentation, regularization tricks, and AutoML approaches. During my PhD, I also spent some time at Google in Prof. Cordelia Schmid's group.
I feel fortunate to have worked on a broad range of computer vision tasks, and I am always looking for new challenging projects and ideas to work on. If you are interested in collaborations or discussing new ideas, feel free to contact me here.
When I am not working, you'll find me on the tennis court or in a bouldering hall :).
PS: I am looking for new opportunities (full-time or consulting) at the moment. You can find a link to my CV here.
Publications
Towards Understanding Adversarial Robustness of Optical Flow Networks, CVPR 2022
Simon Schrodi, Tonmoy Saikia, Thomas Brox
Improving robustness against common corruptions with frequency biased models, ICCV 2021
Tonmoy Saikia, Cordelia Schmid, Thomas Brox
Multi-headed Neural Ensemble Search, ICMLW 2021
Ashwin Narayanan, Arber Zela, Tonmoy Saikia, Thomas Brox, Frank Hutter
Towards improving robustness of compressed CNNs, ICMLW 2021
Jasper Hoffmann, Shashank Agnihotri, Tonmoy Saikia, Thomas Brox
Understanding and robustifying differentiable architecture search, ICLR 2020
Arber Zela, Thomas Elsken, Tonmoy Saikia, Yassine Marrakchi, Thomas Brox, Frank Hutter
AutoDispNet: Improving Disparity Estimation With AutoML, ICCV 2019
Tonmoy Saikia, Yassine Marrakchi, A. Zela, F. Hutter, Thomas Brox
Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation, ECCV 2018
Eddy Ilg, Tonmoy Saikia, Margret Keuper, Thomas Brox
FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks, CVPR 2017
Eddy Ilg, Nikolaus Mayer, Tonmoy Saikia, Margret Keuper, Alexey Dosovitskiy, Thomas Brox
