Md Hasibur Rahman

Research

Reliable learning from incomplete multimodal observations

My research asks how learning systems can make reliable decisions when observations are small, sparse, noisy, asynchronous, or missing. I study this problem through computer vision, spatiotemporal modeling, multimodal sensing, and emerging biomedical applications.

Research agenda

Reliable perception

Detection and tracking of small or weakly observed objects under occlusion, camera motion, missed detections, and uncertain dynamics.

Multimodal sensing

Fusion of heterogeneous evidence across sensors and time, with explicit attention to disagreement, missing measurements, and reliability.

Biomedical and wearable AI

Personalized and multimodal modeling of physiological, behavioral, imaging, and wearable-sensing data, with emphasis on interpretable and clinically meaningful variation.

Representative work

Methods and systems

STARD-Net architecture for spatiotemporal airborne object detection
STARD-Net: spatial and temporal reasoning for tiny airborne targets.

Computer vision · Detection

STARD-Net

A spatiotemporal architecture for detecting tiny airborne objects from moving drones, designed to preserve weak target evidence across frames under clutter, motion, and partial occlusion.

ProjectPaperCode
KRAfT formation-aware multi-object tracking overview
KRAfT: formation-aware association and conservative track recovery.

Computer vision · Tracking

KRAfT

A formation-aware tracking framework that uses residual motion refinement and neighboring-track structure to preserve identity through missed detections and ambiguous associations.

ProjectPaperCode
MultiLiDAR UAV sensing methodology with heterogeneous sensor streams
Multi-LiDAR sensing: aligning sparse, asynchronous evidence over time.

Multimodal AI · Sensor fusion

Track-Aware Reliability Fusion (TARF)

Reliability-aware fusion of sparse, heterogeneous LiDAR streams for 3D drone tracking under asynchronous sampling and sensor disagreement.

ProjectSIGSPATIAL listing
Radiogenomic brain tumor classification problem illustrated with MRI sequences
Radiogenomic classification from complementary MRI sequences.

Biomedical AI · Medical imaging

Radiogenomic MRI classification

An exploration of complementary MRI sequences for non-invasive tumor molecular-status prediction, motivating broader work on robust and interpretable multimodal biomedical learning.

ProjectCode

Additional work in genomic AI, remote sensing, and applied machine learning is listed on the projects page.