NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:75
Title:Social-BiGAT: Multimodal Trajectory Forecasting using Bicycle-GAN and Graph Attention Networks

The authors have presented a strong rebuttal and improved the clarity of the manuscript with the responses to the suggestions of the reviewers. This paper opens up a new avenue for graph neural network researchers, with a clearly motivated architecture which achieves solid results, both quantitatively and, following the rebuttal response, also qualitatively. In discussion with the reviewers regarding the paper and the rebuttal, the main concern remained the lack of a video demo, which would greatly improve the quality of this manuscript. However, considering the strong performance numbers presented in the paper, it would be surprising if the qualitative result in the demo video turns out to be terrible or does not make any sense. Also, the authors included convincing qualitative examples in the paper and the rebuttal already, which kind of addressed the concerns of the major reviewers. We also agreed that notational issues can be fixed through minor editing.