Attentional Bottleneck: Towards an Interpretable Deep Driving Network

TL;DR

This paper introduces the Attentional Bottleneck architecture for deep driving networks, enhancing interpretability by combining visual attention with an information bottleneck, which improves transparency without sacrificing accuracy.

Contribution

The novel Attentional Bottleneck architecture integrates visual attention and information bottleneck to improve interpretability in deep driving models.

Findings

Provides sparse, interpretable attention maps

Maintains or slightly improves model accuracy

Outperforms traditional visual attention models

Abstract

Deep neural networks are a key component of behavior prediction and motion generation for self-driving cars. One of their main drawbacks is a lack of transparency: they should provide easy to interpret rationales for what triggers certain behaviors. We propose an architecture called Attentional Bottleneck with the goal of improving transparency. Our key idea is to combine visual attention, which identifies what aspects of the input the model is using, with an information bottleneck that enables the model to only use aspects of the input which are important. This not only provides sparse and interpretable attention maps (e.g. focusing only on specific vehicles in the scene), but it adds this transparency at no cost to model accuracy. In fact, we find slight improvements in accuracy when applying Attentional Bottleneck to the ChauffeurNet model, whereas we find that the accuracy deteriorates with a traditional visual attention model.