Saksham SuriView profile
Researcher
Saksham Suri is a Researcher at Meta Reality Labs , focusing on efficient foundation models and computer vision . He earned his PhD in Computer Science from the University of Maryland, College Park , advised by Prof. Abhinav Shrivastava . Prior to this, he completed his B.Tech in Computer Science and Engineering at IIIT Delhi , where he worked at the IAB Lab and Precog research group. His research interests center around computer vision and machine learning , particularly in reducing supervision requirements and leveraging synthetic data (e.g., from diffusion models and GANs ) for tasks like object detection , image segmentation , lip synchronization , and video instance segmentation . He has proposed novel architectures and loss functions to enhance recognition through generation, addressing challenges in sparsely annotated object detection , autoregressive generative models , and efficient feature transforms for Vision Transformers . Notable publications include LARP (ICLR 2025, under review), UVIS (CVPR Workshop 2024), Gen2Det (CVPR Workshop 2024), LiFT (ECCV 2024), GRIT (WACV 2024), and Diff2Lip (WACV 2024). These works span topics in video tokenization , synthetic data for detection , lightweight feature transforms , and audio-conditioned lip synchronization . He has collaborated with researchers such as Rama Chellappa , Mayank Vatsa , Richa Singh , C.-C. Jay Kuo , and Larry Davis . His academic contributions include open-source code repositories for projects like LARP , SparseDet , ViT Analysis , and LiFT . He has also served as an Area Chair for CVPR (2018–2021) and contributed to educational resources for courses like CS188 at UC Berkeley.






