most citedSEVD: Synthetic Event-based Vision Dataset for Ego and Fixed Traffic Perception

3 citations · 3 across the 4 of their papers we have counts for

collaborators

4 papers