Papers › Exploring Efficiency of Vision Transformers for Self-Supervised Monocular Depth Estimation

Exploring Efficiency of Vision Transformers for Self-Supervised Monocular Depth Estimation

27 Dec 2022ISMAR 2022 12archive 2025-07-28

Aleksei Karpov, Ilya Makarov

Depth estimation is a crucial task for the creation of depth maps, one of the most important components for augmented reality (AR) and other applications. However, the most widely used hardware for AR and smartphones has only sparse depth sensors with different ground truth depth acquisition methods. Thus, depth estimation models that are robust for downstream AR tasks performance can only be trained reliably using self-supervised learning based on camera information. Previous works in the field mostly focus on self-supervised models with pure convolutional architectures, without taking global spatial context into account.In this paper, we utilize vision transformer architectures for self-supervised monocular depth estimation and propose VTDepth, a vision transformer-based model, which provides a solution to the problem of the global spatial context. We compare various combinations of convolutional and transformer architectures for self-supervised depth estimation and show that the best combination of models is an encoder with a transformer basis and convolutional decoder. Our experiments demonstrate the efficiency of VTDepth for self-supervised depth estimation. Our set of models achieves state-of-the-art performance for self-supervised learning on NYUv2 and KITTI datasets. Our code is available at https://github.com/ahbpp/VTDepth.

PaperPDFCode

Code

ahbpp/VTDepth mentioned in paperpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

DecoderMonocular Depth EstimationSelf-Supervised Learning

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (stereo supervision) Delta < 1.25 0.904 #27 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (stereo supervision) Delta < 1.25^2 0.965 #27 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (stereo supervision) Delta < 1.25^3 0.983 #27 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (stereo supervision) RMSE 4.439 #27 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (stereo supervision) RMSE log 0.178 #27 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (stereo supervision) Sq Rel 0.743 #27 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (stereo supervision) absolute relative error 0.099 #27 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (monocular supervision) Delta < 1.25 0.893 #39 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (monocular supervision) Delta < 1.25^2 0.964 #39 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (monocular supervision) Delta < 1.25^3 0.983 #39 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (monocular supervision) RMSE 4.530 #39 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (monocular supervision) RMSE log 0.182 #39 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (monocular supervision) Sq Rel 0.762 #39 of 55 Archive leaderboard report
Monocular Depth Estimation KITTI Eigen split unsupervised VTDepthB2 (monocular supervision) absolute relative error 0.105 #39 of 55 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AttentionDense ConnectionsLayer NormalizationLinear LayerMulti-Head AttentionResidual ConnectionSoftmaxVision Transformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections