AI Paper Atlas

Research map / 3D vision / Depth and stereo

Revisiting Gradient-Based Uncertainty for Monocular Depth Estimation

Julia Hornauer, Amir El-Ghoussani, Vasileios Belagiannis

TPAMI 2025 · 3D vision › Depth and stereo · approach: Self-supervised monocular depth · 12 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
gradient-based uncertainty vs post-processing baselineMonodepth2 Reg-model · KITTI monoAbs Rel AUSE: 0.044 -> 0.029Improved
gradient-based uncertainty vs inference-only dropout RMSEMonodepth2 Reg-model · KITTI monoRMSE AUSE: 0.854 -> 0.533Improved
add gradient uncertainty on top of Log-modelMonodepth2 Log-model · KITTI monoAbs Rel AUSE: 0.039 -> 0.026Improved
add gradient uncertainty on top of Self-modelMonodepth2 Self-model · KITTI monoAbs Rel AUSE: 0.030 -> 0.024Improved
gradient uncertainty vs Log, nUCEMonodepth2 Log-model · KITTI mononUCE: 0.317 -> 0.006Improved
feature-space flip instead of image flip for reference (ablation)Monodepth2 · KITTI monoAbs Rel AUSE: 0.029 -> 0.028Improved
rotation by 20 degrees instead of flip for reference (ablation)Monodepth2 · KITTI monoAbs Rel AUSE: 0.029 -> 0.033Hurt
diffusion augmentation instead of flip (ablation)MonoViT · KITTI monoAbs Rel AUSE: 0.028 -> 0.034Hurt
feature flip instead of image flip (ablation)MonoViT · KITTI monoAbs Rel AUSE: 0.028 -> 0.026Improved
rotation 20 degrees instead of flip (ablation)Monodepth2 · NYUAbs Rel AUSE: 0.062 -> 0.077Hurt
gradient method vs inference-only dropout on MonoViT stereo, Log-modelMonoViT stereo Log-model · KITTI stereoAbs Rel AUSE: 0.022 -> 0.026Hurt
extra backward pass cost vs dropout inference timeOurs · KITTI Monodepth2Inference time ms: 53.01 -> 18.84Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy