AI Paper Atlas

Research map / 3D vision / Depth and stereo

Boosting Monocular Metric Depth Estimation via Bokeh Rendering

Hangwei Zhang, Armando Fortes, Tianyi Wei, Xingang Pan

ICML 2026 · 3D vision › Depth and stereo · approach: Self-supervised monocular depth · 0 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
add Stage-1 bokeh stack with DSFADepth Anything V2-L · KITTI Eigenδ1: 0.914 -> 0.943Improved
add Stage-1 bokeh stack with DSFAUniDepthV2-L · HAMMER zero-shotδ1: 0.645 -> 0.895Improved
add DSFA on DAv2 on SintelDepth Anything V2-L · Sintel zero-shotδ1: 0.538 -> 0.542No clear effect
add DSFA on UniDepthV2UniDepthV2-L · Sintel zero-shotAbs Rel: 0.496 -> 0.391Improved
add DSFA on DAv2Depth Anything V2-L · Sintel zero-shotAbs Rel: 0.569 -> 0.413Mixed
remove Focus branch from DSFA (ablation)BokehDepth · KITTI Eigenδ1: 0.620 -> 0.943Improved
remove Space branch from DSFA (ablation)BokehDepth · KITTI Eigenδ1: 0.567 -> 0.943Improved
add FiLM conditioning on K (ablation)BokehDepth · KITTI Eigenδ1: 0.602 -> 0.943Improved
use real calibrated bokeh stack instead of repeated sharp image (ablation)BokehDepth · KITTI Eigenδ1: 0.855 -> 0.943Improved
use Stage-1 stacks instead of BokehMe with LiDAR depth (ablation)BokehDepth · KITTI Eigenδ1: 0.914 -> 0.943Improved
use Stage-1 stacks instead of raw FLUX.1-Kontext bokeh (ablation)BokehDepth · KITTI Eigenδ1: 0.608 -> 0.943Improved
use predicted depth rather than sparse LiDAR in BokehMe stacks (ablation)BokehDepth · KITTI Eigenδ1: 0.914 -> 0.918No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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