Kanchana Ranasinghe
Before joining, he was a Machine Learning Engineer at VeracityAI. He went on to join Stony Brook University as a PhD student with Dr. Michael S. Ryoo. He published 1xICCV, 1xNeurIPS, 1xICLR, 1xCVPR papers with us.
Stony Brook University
Kanchana Ranasinghe designed a novel loss function “Orthogonal Projection Loss” that enforces orthogonality in the feature space and enhances the performance & robustness of neural networks. He also participated in other research projects related to Vision Transformers.
Research Outputs
- Orthogonal Projection Loss [ICCV, 2021]
- Intriguing Properties of Vision Transformers [NeurIPS, 2021 (Spotlight)]
- On Improving Adversarial Transferability of Vision Transformers [ICLR, 2022 (Spotlight)]
- Self-supervised Video Transformer [CVPR, 2023 (Oral)]