AI Paper Atlas

Research map / Video / Video segmentation and tracking

Match-Stereo-Videos: Bidirectional Alignment for Consistent Dynamic Stereo Matching

Junpeng Jing, Ye Mao, Krystian Mikolajczyk

ECCV 2024 · Video › Video segmentation and tracking · approach: Learned optical flow estimation · 14 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
replace 2D conv updater with 3D conv updater (ablation)per-frame baseline · Sintel finalTEPE: 1.74 -> 1.34Improved
replace 2D conv with 3D conv updater (ablation)per-frame baseline · Dynamic Replicadelta_1px (accuracy): 2.40 -> 3.38Hurt
enlarge updater kernel to 1x1x15 horizontally (ablation)3D conv updater · Dynamic ReplicaTEPE: 0.082 -> 0.079Improved
use separated instead of shared weights for update modules (ablation)BiDAStereo updater · Sintel finalTEPE: 1.33 -> 1.34No clear effect
add bidirectional alignment with single left frame in correlation (ablation)BiDAStereo · Dynamic ReplicaTEPE: 0.079 -> 0.067Improved
use multiple left reference frames (multi-multi) instead of single left frame (ablation)triple-frame correlation · Dynamic ReplicaTEPE: 0.067 -> 0.082Hurt
share motion hidden state across scales instead of re-initializing (ablation)motion-propagation recurrent unit · Dynamic ReplicaTEPE: 0.511 -> 0.062Improved
increase inference frames from 10 to 20 (ablation)BiDAStereo · Dynamic ReplicaTEPE: 0.081 -> 0.062Improved
increase inference iterations from 20 to 32 (ablation)BiDAStereo · Sintel finalTEPE: 1.24 -> 1.23No clear effect
add Dynamic Replica to SceneFlow training dataBiDAStereo · Sintel finalTEPE: 1.26 -> 1.22Improved
use BiDAStereo instead of CREStereo on Sintel cleanstereo matching · Sintel cleanTEPE: 0.67 -> 0.75Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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