Event-based vision
Representations, recognition, detection, segmentation, tracking, and temporal learning from asynchronous visual data.
Research platform
Neuromorphic Vision Based Human Robot Teaming
A research hub for event-based perception, human action understanding, sim-to-real transfer, robotic manipulation, and safe interaction with humans in dynamic environments.
Event-drivenLow-latency perception
Human-awareAction and state understanding
Robot-readyPerception for interaction
Research direction
Perception that reacts at the speed of interaction
We investigate how event cameras and complementary sensing can make robots more responsive, robust, privacy-aware, and capable of understanding people during close interaction.
Why this matters
Frame-based vision can struggle with motion blur, changing illumination, fixed sampling rates, and unnecessary data redundancy. Event cameras report brightness changes asynchronously, creating a natural fit for rapid human motion, dynamic scenes, and low-latency robotic perception.
The real challenge is not simply collecting events. It is converting sparse event streams into reliable knowledge about human actions, falls, objects, grasp locations, intent, and safety conditions that a robot can use.
Low latencyReduce delay between scene change, human action, perception, and robot response.
Robust transferBridge synthetic events, heterogeneous datasets, and real event-camera deployment.
Human awarenessEstimate actions, gestures, pose, attention, falls, and interaction context.
Efficient deploymentDevelop compact, privacy-preserving, and edge-ready learning systems.
Core themes
Research themes
Five connected areas organize the platform and keep the projects part of one coherent research program.
Transfer and generalisation
Cross-dataset, synthetic-to-real, and hierarchical transfer across sensors, environments, and tasks.
Efficient learning
Lightweight models, quantisation, federated learning, and edge-oriented deployment.
Human-in-the-loop
Human state understanding for shared workspaces, healthcare, assistive systems, and collaboration.
Robotic manipulation
Event-driven object recognition, grasp detection, and perception for dynamic robot interaction.
Research portfolio
Projects
The portfolio contains nine accepted or published projects.
Event Cameras in Human-in-the-Loop Robotics
A systematic review of event-based human state understanding for pose, action, gesture, expression, gaze, safety, and collaborative robotics.
View project →Event-Based Object Recognition and Grasp Detection
A transfer-learning-enhanced multi-task model that jointly recognizes objects and detects grasping positions from event-camera data.
View project →MATRIX-HAR: Lightweight Event-Based Action Recognition
Adaptive temporal resolution, motion-guided features, hierarchical transfer learning, and quantisation for resource-constrained human action recognition.
View project →CurFed-CHARM: Federated Event-Based Action Recognition
Curriculum sequential federated learning for privacy-preserving action recognition under non-IID and heterogeneous event-camera data.
View project →Robust Event-Based Fall Detection Across Domains
A nine-model benchmark that trains on simulated event datasets and evaluates cross-dataset generalisation on real event-camera fall data.
View project →Trigger-Free Synchronization of RGB and Event Cameras
A coaxial RGB-event platform comparing global clock mapping and per-packet anchoring without dedicated trigger hardware.
View project →Sim-to-Real Human Detection and Segmentation
Synthetic-event training and real-event testing for human detection and instance segmentation in dynamic, occluded industrial and hospital scenes.
View project →Adversarial Robustness for Event-Based Human Detection
A unified evaluation of frame-based and event-specific attacks with localization-aware purification for person detection.
View project →Video-to-Event Converters and Sim-to-Real Transfer
A taxonomy and extended analysis of event simulators, conversion techniques, sensor fidelity, and the residual sim-to-real gap.
View project →Open research
Resources
Datasets and repositories are linked where the associated paper makes them publicly available.
CurFed-CHARM dataset and code
Resources for curriculum sequential federated learning on heterogeneous event-camera action data.
Open repository →Event fall-detection benchmark
Converted event datasets and evaluation material for cross-dataset and simulation-to-real benchmarking.
Open repository →Research demonstrations
Use this area for sensor setup videos, model demonstrations, robotic experiments, and deployment material.
Discuss a demonstration →Research output
Selected publications
Each publication is connected to a project page that summarizes the problem, method, evidence, datasets, and relevance to human-in-the-loop robotics.
2026
Applications of Neuromorphic/Event Camera in Robotics With Human in Loop: A Systematic Review, Datasets, and Challenges
M. H. Zafar, S. K. R. Moosavi, and F. Sanfilippo · IEEE Transactions on Human-Machine Systems, 56(1)
2026
Trigger-Free Synchronization of RGB and Event Cameras
Accepted at the Workshop on Neuromorphic Vision: Advantages and Applications of Event Cameras · ECCV 2026
2026
Bridging the Sim-to-Real Gap in Event-Based Human Perception
Accepted at the Workshop on Neuromorphic Vision: Advantages and Applications of Event Cameras · ECCV 2026
2026
Adversarial Robustness for Event-Based Human Detection
Accepted at the Workshop on Neuromorphic Vision: Advantages and Applications of Event Cameras · ECCV 2026
2026
Video-to-Event Converters and Their Role in Sim-to-Real Transfer
Accepted at the Workshop on Neuromorphic Vision: Advantages and Applications of Event Cameras · ECCV 2026
2025
Neuromorphic Event Camera-Based Object Recognition and Grasping Position Detection Using a Transfer Learning-Enhanced Multi-Task Model
M. H. Zafar, S. K. R. Moosavi, and F. Sanfilippo · IEEE Transactions on Automation Science and Engineering, 22
2026
MATRIX-HAR: Lightweight Temporal Motion-Guided Feature Network With Adaptive Temporal Resolution Based Transfer Learning
M. H. Zafar, S. K. R. Moosavi, and F. Sanfilippo · Expert Systems With Applications, 303, 130628
2026
CurFed-CHARM: Curriculum Sequential Federated Learning With a Channel-Hierarchical Action Recognition Model
M. H. Zafar, S. K. R. Moosavi, and F. Sanfilippo · Computer Vision and Image Understanding, 265, 104696
2026
Towards Robust Neuromorphic/Event Vision-Based Fall Detection: Cross-Dataset Generalisation Benchmarking
M. Z. Masood, M. H. Zafar, S. K. R. Moosavi, F. Demrozi, F. Shaukat, and F. Sanfilippo · Expert Systems With Applications, 325, 132675
Research team
Meet the people behind the work
Professor Filippo Sanfilippo leads the project together with Professor Nadia Saad Noori, with Muhammad Hamza Zafar and Syed Kumayl Raza Moosavi contributing as PhD Researchers.