Sarah Ostadabbas
Data-efficient computer vision and machine learning professor
Boston, Massachusetts
Summary
Researcher focused on data-efficient machine learning and computer vision, emphasizing 'learning more from less' for motion-centric video understanding and small-data domains. northeastern+1
Translational and healthcare-oriented work applying computer vision and ML to medical problems such as infant motor function analysis, in-bed pose estimation, stroke rehabilitation, and affective/physiological state monitoring. nih+1
Academic leader and lab director with active roles in mentorship, institutional leadership (Director of Women in Engineering, lab director, center co-director), and community service through organizing and chairing major conferences and workshops. northeastern+1
Prolific scholar with a substantial publication record, patent activity, and recognized awards and grants (including an NSF CAREER award and multiple industry and government research awards). google+1
Work
Education
Projects
Writing
Interpreting face inference models using hierarchical network dis-section
January 1, 2022Work on interpreting face inference models through hierarchical network dissection, published in International Journal of Computer Vision (IJCV) / available on arXiv.
In-Bed Pose Estimation: Deep Learning With Shallow Dataset
January 1, 2019Paper presenting camera-based methods and infrared-selective imaging for in-bed human pose estimation, addressing lighting variations and perspective differences and demonstrating fine-tuning of pre-trained CNNs for shallow in-bed datasets.
Selected recent conference and journal publications (representative)
Multiple peer-reviewed articles on data-efficient learning, motion-centric video understanding, infant action modeling, multi-person tracking, and generative modeling for driving scenes and robotics (selected conference proceedings and arXiv preprints).