Scalable Event Cloud Network for Event-based Classification
Researchers have introduced SECNet, a novel network architecture designed for event-based classification tasks. This new approach utilizes an "Event Cloud" representation, integrating polarity at a structural level to better capture fine-grained temporal information. SECNet also employs feature extraction in the frequency domain to manage computational load and abstract spatio-temporal features effectively. Experiments across ten datasets demonstrate SECNet's scalability, effectiveness, and efficiency. AI
IMPACT Introduces a more efficient and scalable method for processing event camera data, potentially improving real-time applications.