# Evaluating Scalable Bayesian Deep Learning Methods for Robust Computer Vision

## TL;DR

This paper presents a comprehensive evaluation framework for scalable Bayesian deep learning methods in computer vision, comparing ensembling and MC-dropout, and finds ensembling to be more reliable for uncertainty estimation.

## Contribution

The paper introduces a new evaluation framework for epistemic uncertainty estimation in deep learning, specifically designed for real-world computer vision applications, and provides the first extensive comparison of ensembling and MC-dropout.

## Key findings

- Ensembling provides more reliable uncertainty estimates.
- The evaluation framework tests robustness in real-world scenarios.
- Ensembling outperforms MC-dropout in the comparison.

## Abstract

While deep neural networks have become the go-to approach in computer vision, the vast majority of these models fail to properly capture the uncertainty inherent in their predictions. Estimating this predictive uncertainty can be crucial, for example in automotive applications. In Bayesian deep learning, predictive uncertainty is commonly decomposed into the distinct types of aleatoric and epistemic uncertainty. The former can be estimated by letting a neural network output the parameters of a certain probability distribution. Epistemic uncertainty estimation is a more challenging problem, and while different scalable methods recently have emerged, no extensive comparison has been performed in a real-world setting. We therefore accept this task and propose a comprehensive evaluation framework for scalable epistemic uncertainty estimation methods in deep learning. Our proposed framework is specifically designed to test the robustness required in real-world computer vision applications. We also apply this framework to provide the first properly extensive and conclusive comparison of the two current state-of-the-art scalable methods: ensembling and MC-dropout. Our comparison demonstrates that ensembling consistently provides more reliable and practically useful uncertainty estimates. Code is available at https://github.com/fregu856/evaluating_bdl.