AI Paper Atlas

Research map / 3D vision / Depth and stereo

Why Having 10,000 Parameters in Your Camera Model Is Better Than Twelve

Thomas Schops, Viktor Larsson, Marc Pollefeys, Torsten Sattler

CVPR 2020 · 3D vision › Depth and stereo · approach: Camera model and calibration for SfM

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
Use a generic camera model instead of a parametric modelCamera calibration · All tested camerasmedian reprojection errorImproved
Use a non-central generic model instead of a central generic modelGeneric camera calibration · All tested camerasreprojection error and bias metricsImproved
Replace the two parametric models with central radial distortion modelingCamera calibration · Tested camerascalibration accuracyMixed
Increase star-pattern segment count from checkerboard-style 4 segments to a better-tuned count (ablation)Calibration pattern · D435-C and D435-Imedian reprojection errorImproved
Use 16-segment star patterns rather than checkerboard or deltille patterns (ablation)Calibration pattern · D435-C and D435-Imedian reprojection errorImproved
Use intensity-based or X/Y-gradient feature refinement rather than the tested alternatives (ablation)Feature refinement · Camera calibrationmedian reprojection errorImproved
Use X/Y gradients rather than intensities for feature refinement (ablation)Feature refinement · Monochrome camerasmedian reprojection errorImproved
Use intensities rather than X/Y gradients for feature refinement (ablation)Feature refinement · Color camerasmedian reprojection errorImproved
Use a parametric calibration rather than a generic calibrationStereo depth estimation · D435 and Structure Core, roughly 2 m depthdepth discrepancy: about half a centimeter; more than 1 cmHurt
Use Thin-Prism Fisheye rather than central-generic calibrationCamera pose estimation · D435-C, D435-I, SC-C, and SC-Imedian camera-center error: 2.15 mm; 0.25 mm; 1.80 mm; 0.76 mmHurt
Use Thin-Prism Fisheye rather than non-central generic calibration in bundle adjustmentStructure-from-Motion · D435-I and SC-C videosrelative translation error: 0.2% ± 0.1%; 2.9% ± 2.3%Hurt

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