Auxiliary Tasks in Multi-task Learning

TL;DR

This paper demonstrates that incorporating auxiliary tasks into multi-task CNNs for road scene understanding improves performance and training efficiency, supported by experiments on a synthetic dataset from GTA V.

Contribution

It introduces a multi-modal dataset for multi-task road scene understanding and shows that auxiliary tasks enhance CNN performance and training speed.

Findings

Auxiliary tasks boost network accuracy.

Auxiliary tasks reduce training time.

Synthetic dataset enables multi-task learning experiments.

Abstract

Multi-task convolutional neural networks (CNNs) have shown impressive results for certain combinations of tasks, such as single-image depth estimation (SIDE) and semantic segmentation. This is achieved by pushing the network towards learning a robust representation that generalizes well to different atomic tasks. We extend this concept by adding auxiliary tasks, which are of minor relevance for the application, to the set of learned tasks. As a kind of additional regularization, they are expected to boost the performance of the ultimately desired main tasks. To study the proposed approach, we picked vision-based road scene understanding (RSU) as an exemplary application. Since multi-task learning requires specialized datasets, particularly when using extensive sets of tasks, we provide a multi-modal dataset for multi-task RSU, called synMT. More than 2.5 $\cdot$ 10^5 synthetic images, annotated with 21 different labels, were acquired from the video game Grand Theft Auto V (GTA V). Our proposed deep multi-task CNN architecture was trained on various combination of tasks using synMT. The experiments confirmed that auxiliary tasks can indeed boost network performance, both in terms of final results and training time.