AI Paper Atlas

Research map / Video / Video segmentation and tracking

Temporal Event Stereo via Joint Learning with Stereoscopic Flow

Hoonhee Cho, Jae-Young Kang, Kuk-Jin Yoon

ECCV 2024 · Video › Video segmentation and tracking · approach: Event-based video reconstruction · 20 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 feature warping via stereoscopic flow (ablation)single-frame event stereo · MVSEC split 1mean depth error (cm): 15.2 -> 14.7Improved
add cost volume warping on top of feature warping (ablation)temporal event stereo · MVSEC split 1mean depth error (cm): 14.7 -> 13.6Improved
use both feature and cost volume warping (ablation)temporal event stereo · MVSEC split 1mean depth error (cm): 15.2 -> 13.0Improved
use entropy reliability filter instead of addition for cost fusion (ablation)cost volume fusion · MVSEC split 1mean depth error (cm): 14.0 -> 13.0Improved
share stereo features in a lightweight flow network instead of separate EV-Flow (ablation)stereoscopic flow network · MVSEC split 1mean depth error (cm): 13.7 -> 13.0Improved
use hard epipolar constraint (shared vertical flow) instead of soft (ablation)stereoscopic flow · MVSEC split 1mean depth error (cm): 13.3 -> 13.0Improved
use temporal disparity consistency loss instead of contrast maximization only (ablation)flow loss · MVSEC split 1mean depth error (cm): 13.6 -> 13.1Improved
add contrast loss to TDC loss (ablation)flow loss · MVSEC split 1mean depth error (cm): 13.1 -> 13.0Improved
increase voxel grid bins from 1 to 5 (ablation)temporal event stereo · MVSEC split 1mean depth error (cm): 13.7 -> 13.0Improved
increase voxel grid bins from 5 to 10 (ablation)temporal event stereo · MVSEC split 1mean depth error (cm): 13.0 -> 13.7Hurt
use more frames at test time (streaming) than the 4 used in training (ablation)temporal event stereo · MVSEC split 1mean depth error (cm): 13.0 -> 12.8Improved
replace recurrent temporal aggregation with flow-based warpingDTC-SPADE · MVSEC inference costFPS: 11.6 -> 71.6Improved

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