TSANet: Forecasting traffic congestion patterns from aerial videos using graphs and transformers

K. Naveen Kumar, Debaditya Roy, Thakur Ashutosh Suman, Chalavadi Vishnu, C. Krishna Mohan

Abstract: Highlights•Traffic states to quantify congestion in lane-less traffic using spatial graphs.•Transformer TSANet to generate features for future traffic and predict traffic state.•EyeonTraffic — A novel and large annotated aerial dataset for lane-less traffic.