AI Paper Atlas

Research map / 3D vision / Depth and stereo

SaccadeCam: Adaptive Visual Attention for Monocular Depth Sensing

Brevin Tilmon, Sanjeev J. Koppal

ICCV 2021 · 3D vision › Depth and stereo · approach: Self-supervised monocular depth · 6 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 fovea weighting of the photometric loss to SaccadeCam training (ablation)SaccadeCam · KITTI depth, bandwidth (c) target 8 px/mmRMSE: 6.815 -> 6.555Improved
add fovea weighting of the photometric loss to SaccadeCam training (ablation)SaccadeCam · KITTI depth, bandwidth (a) target 35 px/mmAbs Rel: 0.115 -> 0.116No clear effect
add fovea weighting of the photometric loss to SaccadeCam training (ablation)SaccadeCam · KITTI depth, bandwidth (b) target 27 px/mmAbs Rel: 0.121 -> 0.119Improved
learn adaptive resolution distribution (SaccadeCam) instead of uniform target resolutionself-supervised depth network · KITTI depth, bandwidth (c) target 8 px/mmAbs Rel: 0.194 -> 0.164Improved
learn adaptive resolution distribution (SaccadeCam) instead of uniform target resolutionself-supervised depth network · KITTI depth, bandwidth (a) target 35 px/mmAbs Rel: 0.117 -> 0.115Improved
learn adaptive resolution distribution (SaccadeCam) instead of uniform target resolutionself-supervised depth network · KITTI depth, bandwidth (b) target 27 px/mmAbs Rel: 0.118 -> 0.119No clear effect
use learned attention instead of color edge detector as attention proxy (ablation)SaccadeCam · KITTI depth, bandwidth (b) target 27 px/mmAbs Rel: 0.137 -> 0.121Improved
use learned attention instead of color edge detector as attention proxy (ablation)SaccadeCam · KITTI depth, bandwidth (c) target 8 px/mmAbs Rel: 0.167 -> 0.167No clear effect
place focused depth in oracle-selected high-error regions of WAC image (True Oracle)Wide angle camera depth · KITTI oracle study (c)Abs Rel: 0.128 -> 0.112Improved
choose attention regions by photometric error instead of error against LIDAR (True Oracle) (ablation)Oracle SaccadeCam · KITTI oracle study (a)Abs Rel: 0.114 -> 0.116Hurt
increase full resolution to 70 px/mm from target 31 px/mmmonocular depth network · KITTI depth, full resolution 70 px/mmAbs Rel: 0.118 -> 0.109Improved

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