GSTGM: Graph, spatial-temporal attention and generative based model for pedestrian multi-path prediction

Muhammad Haris Kaka Khel, Paul Greaney, Marion McAfee, Sandra Moffett, Kevin Meehan

Abstract: Highlights•This graph model forecasts multiple pedestrian paths, improving safety in crowds.•Our approach uses a GCN and attention, with a generative model for fast convergence.•This approach reduces data requirements and surpasses state-of-the art model accuracy.