Exploring deep learning for Event-Based Saliency Prediction with a Transformer-based model
Researchers have introduced SEST, a novel Transformer-based model for predicting visual saliency from event-based camera data. This work addresses the scarcity of relevant datasets by introducing two new benchmarks, N-DHF1K and N-UCF Sports, generated from existing RGB saliency datasets. SEST demonstrates strong performance, outperforming prior event-based methods and narrowing the gap with state-of-the-art RGB models, while also showing transferability to real-world event camera data. AI
IMPACT Opens a new research direction in event-based vision and neuromorphic visual attention, potentially improving visual processing for specialized cameras.