AI Paper Atlas

Research map / 3D vision / Depth and stereo

Multi-modal Masked Pre-training for Monocular Panoramic Depth Completion

Zhiqiang Yan, Xiang Li, Kun Wang, Zhenyu Zhang, Jun Li, Jian Yang

ECCV 2022 · 3D vision › Depth and stereo · approach: Dense depth prediction and completion · 37 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 shared random masked pre-training (mask size 16, ratio 0.75) (ablation)GuideNet · Stanford2D3DRMSE (mm): 196.7 -> 167.6Improved
use shared mask on RGB-D instead of independent masks (ablation)M3PT · Stanford2D3DRMSE (mm): 182.2 -> 167.6Improved
mask RGB and depth together instead of RGB only (ablation)M3PT · Stanford2D3DRMSE (mm): 194.0 -> 167.6Improved
mask RGB and depth together instead of depth only (ablation)M3PT · Stanford2D3DRMSE (mm): 186.1 -> 167.6Improved
increase mask size from 16 to 32 (ablation)M3PT · Stanford2D3DRMSE (mm): 167.6 -> 169.9Hurt
vary mask size 4, 8, 16 with shared mask (ablation)M3PT · Stanford2D3DRMSE (mm): 166.8, 168.9, 167.6No clear effect
increase mask ratio from 0.15 to 0.75 (size 16) (ablation)M3PT · Stanford2D3DRMSE (mm): 173.5 -> 167.6Improved
extend training without pre-training from 100 to 400 epochs (ablation)GuideNet · Stanford2D3DRMSE (mm): 196.7 -> 196.3No clear effect
pre-train on all three datasets instead of only own data (ablation)M3PT · Stanford2D3DRMSE (mm): 167.6 -> 149.0Improved
pre-train on all three datasets instead of only own data (ablation)M3PT · Matterport3DRMSE (mm): 146.2 -> 138.9Improved
add M3PT pre-trainingS2D · KITTI validationRMSE (mm): 858.02 -> 844.16Improved
add M3PT pre-trainingGuideNet · KITTI validationRMSE (mm): 777.78 -> 761.57Improved

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