AI Paper Atlas

Research map / Video / Video segmentation and tracking

Temporally Consistent Stereo Matching

Jiaxi Zeng, Chengtang Yao, Yuwei Wu, Yunde Jia

ECCV 2024 · Video › Video segmentation and tracking · approach: Layered video decomposition and dynamics · 12 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
initialize refinement from past disparity and hidden state directly (XR-Stereo style) instead of from scratch (ablation)RAFT-Stereo (5 iterations) · TartanAir ablation, all regions>1px (%): 8.04 -> 8.77Hurt
add temporal state fusion of completion-module state with past hidden state (ablation)TC-Stereo · TartanAir ablation, all regions>1px (%): 8.77 -> 6.98Improved
add temporal disparity completion (TDC) module (ablation)TC-Stereo · TartanAir ablation, all regions>1px (%): 6.98 -> 6.10Improved
add dual-space refinement module (ablation)TC-Stereo · TartanAir ablation, occluded regions>1px (%): 29.89 -> 25.89Improved
combine TDC and dual-space refinement (ablation)TC-Stereo · TartanAir ablation, all regions>1px (%): 6.10 -> 5.98Improved
extend training sequence length from 2 to 4 frames (ablation)TC-Stereo · TartanAir ablation, all regions>1px (%): 5.98 -> 5.84Improved
run TC-Stereo in single-frame mode (no temporal inputs at inference) (ablation)TC-Stereo · TartanAir ablation, all regionstemporal jitter |Δd| (px): 0.20 -> 0.28Hurt
use TC-Stereo with 5 iterations instead of RAFT-Stereo with 32 iterationsRAFT-Stereo · TartanAir ablation, all regions>1px (%): 6.02 -> 5.98Improved
increase RAFT-Stereo iterations from 5 to 32 without temporal modules (ablation)RAFT-Stereo · TartanAir ablation, occluded regionstemporal jitter |Δd| (px): 1.15 -> 0.93Improved
use TC-Stereo instead of DLNR on SceneFlow EPETC-Stereo · SceneFlowEPE (px): 0.47 -> 0.50Hurt

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