Kick Back & Relax: Learning to Reconstruct the World by Watching SlowTV

TL;DR

This paper introduces a large-scale SlowTV dataset from YouTube for self-supervised monocular depth estimation, enabling models to generalize across diverse environments and outperform existing approaches.

Contribution

It presents a new extensive dataset and training practices that improve zero-shot generalization and performance of SS-MDE models across varied real-world scenes.

Findings

Model outperforms existing SSL methods

Achieves near-supervised state-of-the-art performance

Demonstrates strong zero-shot generalization

Abstract

Self-supervised monocular depth estimation (SS-MDE) has the potential to scale to vast quantities of data. Unfortunately, existing approaches limit themselves to the automotive domain, resulting in models incapable of generalizing to complex environments such as natural or indoor settings. To address this, we propose a large-scale SlowTV dataset curated from YouTube, containing an order of magnitude more data than existing automotive datasets. SlowTV contains 1.7M images from a rich diversity of environments, such as worldwide seasonal hiking, scenic driving and scuba diving. Using this dataset, we train an SS-MDE model that provides zero-shot generalization to a large collection of indoor/outdoor datasets. The resulting model outperforms all existing SSL approaches and closes the gap on supervised SoTA, despite using a more efficient architecture. We additionally introduce a collection of best-practices to further maximize performance and zero-shot generalization. This includes 1) aspect ratio augmentation, 2) camera intrinsic estimation, 3) support frame randomization and 4) flexible motion estimation. Code is available at https://github.com/jspenmar/slowtv_monodepth.