Vehicle Trajectory Prediction by Transfer Learning of Semi-Supervised Models

TL;DR

This paper demonstrates that semi-supervised learning significantly enhances vehicle trajectory prediction accuracy on real-world benchmarks by leveraging large unlabeled datasets and transfer learning techniques.

Contribution

It introduces the use of semi-supervised models with transfer learning for vehicle trajectory prediction, showing improved performance over supervised models.

Findings

Semi-supervised models outperform supervised models on benchmarks.

Using unlabeled data scales pre-training from millions to a billion images.

Contrastive learning and teacher-student methods are effective within semi-supervised frameworks.

Abstract

In this work we show that semi-supervised models for vehicle trajectory prediction significantly improve performance over supervised models on state-of-the-art real-world benchmarks. Moving from supervised to semi-supervised models allows scaling-up by using unlabeled data, increasing the number of images in pre-training from Millions to a Billion. We perform ablation studies comparing transfer learning of semi-supervised and supervised models while keeping all other factors equal. Within semi-supervised models we compare contrastive learning with teacher-student methods as well as networks predicting a small number of trajectories with networks predicting probabilities over a large trajectory set. Our results using both low-level and mid-level representations of the driving environment demonstrate the applicability of semi-supervised methods for real-world vehicle trajectory prediction.