Bishoy Galoaa
PhD student and computer vision researcher
Summary
Work
Education
Projects
Writing
More Than Meets the Eye: Enhancing Multi-Object Tracking Even with Prolonged Occlusions
January 1, 2025Introduces MOTE, a multi-object tracking algorithm using deformable detection transformers, optical flow, and a disocclusion matrix to maintain track consistency under prolonged occlusions; reports strong results on MOT17/MOT20/DanceTrack.
DragonTrack: Transformer-Enhanced Graphical Multi-Person Tracking in Complex Scenarios
January 1, 2025Presents DragonTrack, an end-to-end framework combining detection transformers and graph convolutional networks for multi-person tracking; demonstrates improved HOTA and MOTA on MOT benchmarks.
Uncertainty-Aware Ankle Exoskeleton Control
January 1, 2025An uncertainty-aware control framework enabling ankle exoskeletons to detect out-of-distribution movements and safely disengage assistance; evaluates ensemble uncertainty estimators and online testing.
Track and Caption Any Motion: Query-Free Motion Discovery and Description in Videos
January 1, 2025Proposes TCAM, a motion-centric framework that discovers and describes motion patterns without queries by aligning motion fields with vision-language representations for spatial grounding.
K-Track: Kalman-Enhanced Tracking for Accelerating Deep Point Trackers on Edge Devices
January 1, 2025Introduces K-Track, a hybrid framework combining sparse deep learning keyframe updates with Kalman filtering to accelerate point trackers for edge deployment while retaining accuracy.
Advancing prognostics in oncology: developing a machine learning model for predicting 2-year and 5-year survival rates in patients with undifferentiated pleomorphic sarcoma
January 1, 2025Clinical machine learning work in orthopedic oncology contributing to prognostic models and related reflections published in Annals of Surgical Oncology / PubMed entries.