Boris N. Oreshkin

Papers

Total Citations

220

H-Index

About

Boris N. Oreshkin is a leading researcher in few-shot learning and computer vision, with a particular focus on enabling deep learning models to perform effectively with minimal labeled data—a critical challenge for robotics and real-world AI applications. His most influential work, "AMP: Adaptive Masked Proxies for Few-Shot Segmentation" (2019), has garnered over 200 citations, establishing him as a key contributor to sample-efficient segmentation. Oreshkin’s major contribution lies in developing the adaptive masked proxies method, which constructs final segmentation layer weights from just a few labeled samples by leveraging multi-resolution average pooling. This innovation directly addresses the limitations of traditional deep learning, which relies on large-scale datasets, and offers a practical solution for domains like robotics where data collection is costly. His research demonstrates how to maintain high performance in segmentation tasks under extreme data scarcity, making his work foundational for advancing autonomous systems. With a total citation count exceeding 220 across his most cited papers, Oreshkin’s impact is evident in the growing adoption of few-shot techniques in vision and robotics, positioning him as a pivotal figure in bridging the gap between data-hungry models and real-world efficiency.