Generation of Pedestrian Crossing Scenarios Using Ped-Cross Generative Adversarial Network

The safety of vulnerable road users is paramount importance as transport moves towards fully automated driving. richness real-world data required for testing autonomous vehicles limited and furthermore, available do not present a fair representation different scenarios rare events. Before deploying publicly, their abilities must reach threshold, least with regards to users, such pedestrians. In this paper, we novel Generative Adversarial Networks named the Ped-Cross GAN. GAN able generate crossing sequences pedestrians in form human pose sequences. trained Pedestrian Scenario dataset. dataset, derived from existing datasets, enables training on richer pedestrian scenarios. We demonstrate an example its use through results show that new are same distribution those contained Having method these capabilities important future transport, it will allow adequate Connected Autonomous Vehicles how they correctly perceive intention street, ultimately leading fewer casualties our roads.