Federico Angelini
Estimating abnormal human activity using multimodal surveillance.
Email: [email protected]
Project title
Multimodal surveillance
Supervisors
Project description
This project focuses on Machine Learning and Fusion Techniques. The techniques will detect human activity using multimodal wide-area sensor measurements.
We will develop a robust model for video-based human activity. The model will estimate abnormal human actions. These will be based on multimodal data and contextual information.
The main techniques that we use include:
- Deep Learning based architectures
- multimodal data fusion
- video data augmentation
The project involves extensive data recordings, software design, simulations and demo realisation.
We have achieved breakthrough results. These have led to:
- academic publications
- dataset recording
- effective deep learning algorithms for human action classification and anomaly detection.
Publications
- Angelini F, Fu Z, Long Y, Shao L, Naqvi SM. 2D Pose-based Real-time Human Action Recognition with Occlusion-handling. IEEE Transactions on Multimedia 2019, epub ahead of print.
- Angelini F, Naqvi SM. Joint RGB-Pose Based Human Action Recognition for Anomaly Detection Applications. In: 22nd International Conference on Information Fusion (FUSION). 2019, Ottawa, Canada.
- Angelini F, Yan J, Naqvi SM. Privacy-preserving Online Human Behviour Anomaly Detection Based on Body Movements and Objects Positions. In: 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2019, Brighton, UK: IEEE.
- Angelini F, FU Z, Chambers JA, Naqvi MN. 3D-HOG Embedding Frameworks for Single and Multi-Viewpoints Action Recognition Based on Human Silhouettes. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2018). 2018, Calgary, Alberta, Canada: IEEE.
Interests
Artificial Intelligence, Deep Learning
Qualifications
- BSc in Mathematics and MSc in Pure and Applied Mathematics, Università degli Studi di Roma “Tor Vergata”