AI Paper Atlas

Research map / 3D vision / Depth and stereo

RADU: Ray-Aligned Depth Update Convolutions for ToF Data Denoising

Michael Schelling, Pedro Hermosilla, Timo Ropinski

CVPR 2022 · 3D vision › Depth and stereo · approach: Neural light field and illumination imaging · 18 citations (OpenAlex)

Check 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 cyclic self-training U-DARADU · S4MAE [cm]: 1.83 -> 2.11Hurt
Add cyclic self-training U-DARADU · S5MAE [cm]: 2.59 -> 1.63Improved
Use RADU instead of TIRToF denoising · S4MAE [cm]: 2.60 -> 1.83Improved
Use RADU instead of CFNToF denoising · Cornell-BoxMAE [cm]: 3.99 -> 3.64Improved
Use RADU instead of CFNToF denoising · FLATMAE [cm]: 6.29 -> 3.31Improved
Replace standard MCConv with MCConv plus ray-aligned depth updates (ablation)Latent 3D block · S3 validationMAE [cm]: 2.51 -> 2.28Improved
Use 2.5D convolutions instead of 2D convolutions (ablation)Latent block · S3 validationMAE [cm]: 2.72 -> 2.61Improved
Remove ray-aligned point updates by setting α=0.0 (ablation)RADU · S3 validationMAE [cm]: 2.28 -> 2.51Hurt
Increase update scale α from 0.1 to 0.2 (ablation)RADU · S3 validationMAE [cm]: 2.28 -> 2.81Hurt
Increase update scale α from 0.1 to 1.0 (ablation)RADU · S3 validationMAE [cm]: 2.28 -> 3.42Hurt
Set update scale α to each layer's receptive-field radius (ablation)RADU · S3 validationMAE [cm]: 2.28 -> 2.45Hurt
Remove the intermediate coarse-depth L1 loss (ablation)RADU · S3MAE [cm]: 2.28 -> 2.63; +0.35Hurt

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