A UAV-oriented sensing pipeline built for detection tasks where small targets, weak contrast, and field constraints make single-sensor approaches unreliable.

What I built

  • Fusion strategies across EO, IR, GPR, and magnetometer inputs
  • Feature-level, decision-level, and probabilistic late-fusion methods
  • Real-time inference workflows designed for operational environments
  • Evaluation patterns tuned for low-contrast targets and imbalanced detection conditions

Outcome

  • Improved robustness compared with isolated sensor workflows
  • Supported field-ready deployment requirements rather than lab-only experimentation
  • Demonstrated how multimodal sensing can recover signal where single modalities underperform

Stack

Computer vision, multimodal fusion, real-time inference, UAV sensing, and detection systems