In this paper, we propose HOME, a framework tackling the motion forecasting problem with an image output representing the probability distribution of the agents future location. This method allows for a simple architecture with classic convolution networks coupled with attention mechanism for agent interactions, and outputs an unconstrained 2D top-view representation of the agents possible future. Based on this output, we design two methods to sample a finite set of agents future locations. These methods allow us to control the optimization trade-off between miss rate and final displacement error for multiple modalities without having to retrain any part of the model. We apply our method to the Argoverse Motion Forecasting Benchmark and achieve 1st place on the online leaderboard.
Download