AI Paper Atlas

Research map / Video / Video segmentation and tracking

SeLFVi: Self-Supervised Light-Field Video Reconstruction From Stereo Video

Prasan Shedligeri, Florian Schiffers, Sushobhan Ghosh, Oliver Cossairt, Kaushik Mitra

ICCV 2021 · Video › Video segmentation and tracking · approach: Video restoration with temporal priors · 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 epipolar geometric consistency loss L_geo (ablation)SeLFVi · LF video ablationPSNR: 19.20 -> 32.39Improved
add stereo photometric loss L_stereo (ablation)SeLFVi · LF video ablationPSNR: 32.20 -> 32.39No clear effect
add temporal loss L_temp (ablation)SeLFVi · LF video ablationPSNR: 31.98 -> 32.39No clear effect
add tensor-display (low-rank) layer to the model (ablation)SeLFVi · LF video ablationPSNR: 30.50 -> 32.39Improved
add tensor-display layer when L_geo is absent (ablation)SeLFVi · LF video ablationPSNR: 6.04 -> 19.20Improved
increase rank M of F from 1 to 12 (ablation)SeLFVi · 7x7 LF outputPSNR: 31.43 -> 32.39Improved
increase rank M of F from 1 to 12 (ablation)SeLFVi · 7x7 LF outputinference time: 0.103 -> 0.381Hurt
use direct regression instead of low-rank intermediate representation (ablation)SeLFVi · 7x7 LF outputinference time: 0.381 -> 0.108Improved
use SeLFVi instead of X-fields 2-viewLF video from stereo · Average of three datasetsPSNR: 27.12 -> 33.29Improved
use SeLFVi instead of X-fields 4-view (which needs four views)LF video from stereo · Average of three datasetsPSNR: 30.84 -> 33.29Improved
use SeLFVi instead of HighRes warpingLF video from stereo · Average of three datasetsLPIPS: 0.063 -> 0.071Hurt
use SeLFVi instead of X-fields 4-viewLF video from stereo · Temporal consistencywarping MAE: 1.73 -> 1.58Improved

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