Why ADE and FDE are not the best metrics to score motion prediction

<p>The most common metrics are the average displacement error (ADE - mean L2 distance between all trajectory points) and final displacement error (FDE - L2 distance between final trajectory points).</p> <p>Those metrics are easy to compute and relatively easy to interpret, but they have some potential drawbacks that I want to talk about in this blog post. First, let&#39;s look at the quality of a predicted trajectory from the planer perspective.</p> <p><a href="https://towardsdatascience.com/why-ade-and-fde-might-not-be-the-best-metrics-to-score-motion-prediction-model-performance-and-what-1980366d37be"><strong>Website</strong></a></p> <p>&nbsp;</p>
Tags: ADE FDE