Pedestrian trajectory prediction with convolutional neural networks

Predicting the future trajectories of pedestrians is a challenging problem that has range application, from crowd surveillance to autonomous driving. In literature, methods approach pedestrian trajectory prediction have evolved, transitioning physics-based models data-driven based on recurrent neural networks. this work, we propose new prediction, with introduction novel 2D convolutional model. This model outperforms models, and it achieves state-of-the-art results ETH TrajNet datasets. We also present an effective system represent positions powerful data augmentation techniques, such as addition Gaussian noise use random rotations, which can be applied any As additional exploratory analysis, experimental inclusion occupancy social information, empirically show these are ineffective in capturing interaction.