Back to MLP: A Simple Baseline for Human Motion Prediction

TL;DR

This paper demonstrates that a simple, lightweight MLP-based model, combined with standard practices like DCT and residual prediction, can outperform complex state-of-the-art methods in human motion prediction.

Contribution

The authors introduce siMLPe, a minimal MLP-based approach that surpasses complex models by leveraging standard techniques and residual prediction for human motion forecasting.

Findings

siMLPe outperforms state-of-the-art methods on multiple datasets

A lightweight MLP with 0.14 million parameters is sufficient for high performance

Standard practices like DCT and residual prediction significantly improve results

Abstract

This paper tackles the problem of human motion prediction, consisting in forecasting future body poses from historically observed sequences. State-of-the-art approaches provide good results, however, they rely on deep learning architectures of arbitrary complexity, such as Recurrent Neural Networks(RNN), Transformers or Graph Convolutional Networks(GCN), typically requiring multiple training stages and more than 2 million parameters. In this paper, we show that, after combining with a series of standard practices, such as applying Discrete Cosine Transform(DCT), predicting residual displacement of joints and optimizing velocity as an auxiliary loss, a light-weight network based on multi-layer perceptrons(MLPs) with only 0.14 million parameters can surpass the state-of-the-art performance. An exhaustive evaluation on the Human3.6M, AMASS, and 3DPW datasets shows that our method, named siMLPe, consistently outperforms all other approaches. We hope that our simple method could serve as a strong baseline for the community and allow re-thinking of the human motion prediction problem. The code is publicly available at \url{https://github.com/dulucas/siMLPe}.