ETA: Energy-based Test-time Adaptation for Depth Completion

TL;DR

ETA introduces a test-time adaptation method for depth completion models that uses an energy model to detect and adapt to out-of-distribution data without prior knowledge of target environments, improving accuracy.

Contribution

The paper presents ETA, a novel energy-based approach for test-time adaptation of depth completion models using adversarial perturbations to quantify in- or out-of-distribution predictions.

Findings

ETA outperforms previous methods by 6.94% outdoors and 10.23% indoors.

The energy model effectively detects out-of-distribution regions.

Test-time adaptation improves depth prediction accuracy across diverse datasets.

Abstract

We propose a method for test-time adaptation of pretrained depth completion models. Depth completion models, trained on some ``source'' data, often predict erroneous outputs when transferred to ``target'' data captured in novel environmental conditions due to a covariate shift. The crux of our method lies in quantifying the likelihood of depth predictions belonging to the source data distribution. The challenge is in the lack of access to out-of-distribution (target) data prior to deployment. Hence, rather than making assumptions regarding the target distribution, we utilize adversarial perturbations as a mechanism to explore the data space. This enables us to train an energy model that scores local regions of depth predictions as in- or out-of-distribution. We update the parameters of pretrained depth completion models at test time to minimize energy, effectively aligning test-time predictions to those of the source distribution. We call our method ``Energy-based Test-time Adaptation'', or ETA for short. We evaluate our method across three indoor and three outdoor datasets, where ETA improve over the previous state-of-the-art method by an average of 6.94% for outdoors and 10.23% for indoors. Project Page: https://fuzzythecat.github.io/eta.