Random Finite Set framework for anomaly detection

PhD research at RMIT University (2018–2022) on point-pattern learning and Random Finite Set (RFS) theory for anomaly detection.

Key contributions:

• An energy-of-point-pattern approach that treats image features as a point pattern and detects defects through the RFS likelihood of the pattern (EAAI, 2024).
• Deep temporal encoding-decoding networks with multi-head attention for video anomaly detection (Expert Systems with Applications, 2022).
• A multiple-instance learning formulation of the same architecture for weakly labelled video (Expert Systems with Applications, 2023).

Supervisors: Prof. Alireza Bab-Hadiashar and Prof. Reza Hoseinnezhad.