Research vision and interests

Advancing reliable AI systems for vision, recognition, and anomaly detection.

I am a computer vision and machine learning researcher. My work bridges theoretical advancements and practical applications by developing robust pattern recognition and machine learning methods for:

  • Computer vision

  • Biometrics

  • Anomaly detection

  • Time-series analysis

Core research interests: One-class classification, Kernel-based learning, Sparse modelling. My goal is to build systems that perform reliably under limited supervision and in unseen conditions.

Research areas

  • Face Presentation Attack DetectionDetect spoofing attempts such as masks, printed photos, and video replays.

  • Deepfake and Face Manipulation DetectionClassify manipulated content using one-class and open-set learning.

  • One-Class and Open-Set RecognitionDetect and classify when only “normal” data is available during training.

  • Ensemble LearningCombine multiple sources of information at decision, feature, or score levels.

Key research projects

Ongoing and future research directions

Semi-supervised deepfake detection

  • Train detectors with limited labels using semi-supervised learning.

  • Apply one-class and open-set recognition to counter novel attack types.

  • Focus on domain generalisation for robustness across datasets and manipulation methods.

Sparsity-induced classifier fusion

  • Design sparse fusion frameworks to combine complementary classifiers.

  • Implement locally adaptive fusion strategies for spatially variant data.

  • Explore fusion-guided attention mechanisms for data-driven weighting.

Anomaly detection in sequential data streams

  • Extend kernel-based anomaly detection to temporal and streaming contexts with structured embeddings.

  • Develop multi-output kernel regressors for structured predictions in dynamic environments.