Mariia Khan

Papers

Total Citations

H-Index

About

Mariya Khan is a rising researcher in computer vision, with a focus on segmentation, scene understanding, and simulation-based learning for robotics. Her work addresses critical challenges in multi-class multi-instance segmentation, where she developed the Segment Any Object Model (SAOM), introducing a real-to-simulation fine-tuning strategy that enhances the foundational Segment Anything Model (SAM) for more precise object-level segmentation. This work, already garnering early citations, pushes the boundaries of how models handle multiple object classes and instances simultaneously. Khan also contributed to panoramic indoor scene understanding with PanoSCU, a simulation-based dataset designed to support tasks like visual room rearrangement in robotics. By providing a broader spatial context through panoramic views, this dataset enables agents to better understand and restore environments. Her research bridges the gap between simulation and real-world application, offering practical tools for autonomous systems. With her innovative approaches to segmentation and scene understanding, Mariya Khan is establishing herself as a promising voice in the intersection of computer vision and embodied AI.