Aggregated Sparse Attention for Steering Angle Prediction

TL;DR

This paper introduces a novel sparse attention mechanism for visual data in autonomous driving, specifically improving steering angle prediction by aggregating attention, and demonstrates its effectiveness over traditional methods.

Contribution

The paper presents the first application of sparse attention in the visual domain for steering prediction, with an aggregated extension that enhances performance.

Findings

Improved steering angle prediction accuracy.

Sparse attention outperforms no-attention and other attention types.

Aggregation further boosts model performance.

Abstract

In this paper, we apply the attention mechanism to autonomous driving for steering angle prediction. We propose the first model, applying the recently introduced sparse attention mechanism to visual domain, as well as the aggregated extension for this model. We show the improvement of the proposed method, comparing to no attention as well as to different types of attention.