AI Paper Atlas

Research map / 3D vision / Depth and stereo

Learning Depth from Focus in the Wild

Changyeon Won, Hae-Gon Jeon

ECCV 2022 · 3D vision › Depth and stereo · approach: Depth from light field and defocus · 33 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 alignment network (ablation)depth-from-focus network · Rendered focal stacksMAE: 0.0247 -> 0.0151Improved
use learned alignment instead of initial FoV warping (ablation)depth-from-focus network · Rendered focal stacksMAE: 0.0165 -> 0.0151Improved
use alignment network instead of homography (ablation)depth-from-focus network · Rendered focal stacksinference time (s): 0.8708 -> 0.0923Improved
use alignment network instead of homography (ablation)depth-from-focus network · Rendered focal stacksMAE: 0.0151 -> 0.0151No clear effect
use SRD instead of 2D ResNet block (ablation)depth-from-focus network · DefocusNet datasetMAE: 0.0421 -> 0.0403Improved
use SRD instead of 3D ResNet block (ablation)depth-from-focus network · DefocusNet datasetMAE: 0.0409 -> 0.0403Improved
use EFD instead of max-pooling + 3D Conv (ablation)depth-from-focus network · DefocusNet datasetMAE: 0.0421 -> 0.0403Improved
use EFD instead of strided conv (ablation)depth-from-focus network · DefocusNet datasetMAE: 0.0419 -> 0.0403Improved
use EFD instead of 3D pooling layer (ablation)depth-from-focus network · DefocusNet datasetMAE: 0.0414 -> 0.0403Improved
train with random number of focal slices instead of fixed five (ablation)depth-from-focus network · DefocusNet datasetdepth errorImproved
proposed network vs AiFDepthNetdepth-from-focus network · DDFF 12-Scenedelta<1.25 accuracy: 68.33 -> 77.96Improved
proposed network vs AiFDepthNet, cross-datasetdepth-from-focus network · Middlebury (trained on FlyingThings3D)MAE: 3.825 -> 1.645Improved

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