Back to square one: probabilistic trajectory forecasting without bells and whistles

TL;DR

This paper presents a simple yet effective probabilistic neural network model for trajectory forecasting from visual data, achieving competitive results on standard datasets without complex bells and whistles.

Contribution

It introduces a straightforward spatio-temporal CNN for probabilistic trajectory forecasting that performs comparably or better than more complex existing methods.

Findings

Achieves state-of-the-art or comparable results on MNISTseq and Stanford Drones datasets.

Provides explicit probability distributions over trajectory continuations.

Demonstrates effectiveness of a simple model in a complex task.

Abstract

We introduce a spatio-temporal convolutional neural network model for trajectory forecasting from visual sources. Applied in an auto-regressive way it provides an explicit probability distribution over continuations of a given initial trajectory segment. We discuss it in relation to (more complicated) existing work and report on experiments on two standard datasets for trajectory forecasting: MNISTseq and Stanford Drones, achieving results on-par with or better than previous methods.