Unsupervised Pedestrian Pose Prediction: A Deep Predictive Coding Network-Based Approach for Autonomous Vehicle Perception
Xiaoxiao Du · Ram Vasudevan · Matthew Johnson-Roberson
TL;DR
A predictive coding network forecasts pedestrian poses from 2D camera video without labels, tested on two real-world autonomous vehicle data sets.
BRIEF
Self-driving cars and other mobile robots need to anticipate how pedestrians will move, not just where they are now. Predicting body pose a moment ahead supports safer navigation and better human-robot interaction. This work applies a deep predictive coding network (PredNet) to forecast pedestrian poses from ordinary 2D camera images, without labeled pose data for training. The approach is tested on two real-world autonomous vehicle data sets, and the article also outlines future directions for unsupervised and semisupervised pose prediction.
Abstract
Pedestrian pose prediction is an important topic, related closely to robotics and automation. Accurate predictions of human poses and motion can facilitate a more thorough understanding and analysis of human behavior, which benefits real-world applications such as human-robot interaction, humanoid and bipedal robot design, and safe navigation of mobile robots and autonomous vehicles. This article describes a deep predictive coding network (PredNet)-based approach for unsupervised pedestrian pose prediction from 2D camera imagery and provides experimental results of two real-world autonomous vehicle data sets. The article also discusses topics for future work in unsupervised and semisupervised pedestrian pose prediction and its potential applications in robotics and automation systems.