Trajectory Prediction for Heterogeneous Agents: A Performance Analysis on Small and Imbalanced Datasets

The Python implementation of class-conditioned human motion prediction methods, in the paper: Trajectory Prediction for Heterogeneous Agents: A Performance Analysis and on Small and Imbalanced Datasets. 

For human motion prediction, both deep learning-based methods (RED, cRED, TF, cTF, GAN, cGAN, VAE, and cVAE) and MoD-based methods (CLiFF-HMP) are provided.

Class-conditioned motion prediction is an appealing way to reduce forecast uncertainty and get more accurate predictions for heterogeneous agents. However, this is hardly explored in the prior art, especially for mobile robots and in limited data applications. In this paper, different class-conditioned trajectory prediction methods on two datasets are analyzed. A set of conditional pattern-based and efficient deep learning- based baselines are proposed and evaluated on robotics and outdoors datasets (TH ¨OR-MAGNI and Stanford Drone Dataset). The experiments show that all methods improve accuracy in most of the settings when considering class labels. More importantly, there are significant differences when learning from imbalanced datasets, or in new environments where sufficient data is not available. In particular, deep learning methods perform better on balanced datasets, but in applications with limited data, e.g., cold start of a robot in a new environment, or imbalanced classes, pattern-based methods may be preferable.