NeVi 2024
Workshop on Neuromorphic Vision:
Advantages and Applications of Event Cameras
In conjunction with ECCV 2024
Milan
SCOPE and motivation
Neuromorphic sensors, also known as event cameras, are a class of imaging devices mimicking biological visual systems. Unlike traditional frame-based cameras, which capture images synchronously, neuromorphic sensors continuously generate events capturing asynchronous illumination changes.
Event cameras have initially gained interest in the field of robotics due to their low power consumption, extremely low latency, high dynamic range and absence of motion blur. Yet, this wide range of intriguing properties has rapidly enabled new, cutting-edge applications, especially for motion-centric tasks. The very fine temporal granularity of event cameras allows to easily capture complex temporal dynamics in a scene, so that the tackling of complex tasks can abstract from the low-level processing, and focus directly on higher-level cognition.
In the past few years, we have witnessed the development of new astonishing technologies based on neuromorphic vision: low latency and low power consumption have allowed drones to effectively avoid fast-moving obstacles; high dynamic range and lack of motion blur allowed self-driving cars to detect other vehicles and pedestrians in adverse conditions such as low illumination; micro-second temporal granularity has enhanced the analysis of human micro-expressions and emotions. Many other groundbreaking applications are leveraging neuromorphic sensors, from high-speed object counting and defect detection to vibration measurement, fluid monitoring and time-to-contact estimation for spacecraft landing. Event-based processing has also been shown to provide an extra layer of privacy preservation compared to standard cameras, an important addition especially in light of the recent definition of the AI Act by the European Commission to regulate the development of artificial intelligence.
Topics
This workshop aims to foster the growth of event-based research, by gathering researchers in the field and improving the communication between academia and industry, towards the discovery of new bleeding-edge neuromorphic technologies.
Event-based Vision
Representations for event-based data
Event camera simulators
Event-based datasets
Novel sensing techniques for event-based vision
Neuromorphic Event Data Processing
Spiking neural networks
Bio-inspired computational methods
Event-based spatio-temporal feature extraction
Learning methodologies with event data
Neuromorphic vision applications
Event-based human analysis
Driving monitoring systems
Neuromorphic cameras for space
High-speed counting
Autonomous navigation
Hardware architectures for event-based vision
ASIC and FPGA-based implementations
Novel circuitry designs
Benchmarking and characterisation of event-based cameras