Transformers in Self-Supervised Monocular Depth Estimation with Unknown Camera Intrinsics

TL;DR

This paper explores the adaptation of vision transformers for self-supervised monocular depth estimation, demonstrating comparable performance to CNNs with enhanced robustness and generalization, especially when camera intrinsics are unknown.

Contribution

It introduces a method to adapt vision transformers for depth estimation and compares their performance and robustness against CNNs on benchmark datasets.

Findings

Transformers achieve comparable depth estimation accuracy to CNNs.

Transformers exhibit greater robustness to corruptions and adversarial attacks.

Performance remains strong even when camera intrinsics are unknown.

Abstract

The advent of autonomous driving and advanced driver assistance systems necessitates continuous developments in computer vision for 3D scene understanding. Self-supervised monocular depth estimation, a method for pixel-wise distance estimation of objects from a single camera without the use of ground truth labels, is an important task in 3D scene understanding. However, existing methods for this task are limited to convolutional neural network (CNN) architectures. In contrast with CNNs that use localized linear operations and lose feature resolution across the layers, vision transformers process at constant resolution with a global receptive field at every stage. While recent works have compared transformers against their CNN counterparts for tasks such as image classification, no study exists that investigates the impact of using transformers for self-supervised monocular depth estimation. Here, we first demonstrate how to adapt vision transformers for self-supervised monocular depth estimation. Thereafter, we compare the transformer and CNN-based architectures for their performance on KITTI depth prediction benchmarks, as well as their robustness to natural corruptions and adversarial attacks, including when the camera intrinsics are unknown. Our study demonstrates how transformer-based architecture, though lower in run-time efficiency, achieves comparable performance while being more robust and generalizable.