Out-of-Distribution Detection for Monocular Depth Estimation

TL;DR

This paper introduces a novel out-of-distribution detection method for monocular depth estimation that uses reconstruction error to identify OOD images, outperforming existing approaches without altering the depth model.

Contribution

It proposes a reconstruction-error-based OOD detection method for monocular depth estimation that works with fixed models and outperforms current uncertainty estimation techniques.

Findings

High detection accuracy on NYU Depth V2 and KITTI datasets.

Method outperforms existing uncertainty estimation approaches.

Works with different models without retraining the depth estimator.

Abstract

In monocular depth estimation, uncertainty estimation approaches mainly target the data uncertainty introduced by image noise. In contrast to prior work, we address the uncertainty due to lack of knowledge, which is relevant for the detection of data not represented by the training distribution, the so-called out-of-distribution (OOD) data. Motivated by anomaly detection, we propose to detect OOD images from an encoder-decoder depth estimation model based on the reconstruction error. Given the features extracted with the fixed depth encoder, we train an image decoder for image reconstruction using only in-distribution data. Consequently, OOD images result in a high reconstruction error, which we use to distinguish between in- and out-of-distribution samples. We built our experiments on the standard NYU Depth V2 and KITTI benchmarks as in-distribution data. Our post hoc method performs astonishingly well on different models and outperforms existing uncertainty estimation approaches without modifying the trained encoder-decoder depth estimation model.