AI Paper Atlas

Research map / Video / Video segmentation and tracking

BVINet: Unlocking Blind Video Inpainting with Zero Annotations

Zhiliang Wu, Kerui Chen, Kun Li, Hehe Fan, Yi Yang

ICCV 2025 · Video › Video segmentation and tracking · approach: Video restoration with temporal priors · 57 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
use DWT down-sampling in short-term prediction module (ablation)MPNet · mask predictionBCE: 1.1251 -> 1.0785Improved
use DWT down-sampling in short-term prediction module (ablation)MPNet · mask predictionIOU: 0.8437 -> 0.8682Improved
add long-term refinement module (ablation)MPNet DWT_STP · mask predictionBCE: 1.0785 -> 0.9176Improved
use full MPNet (wavelet down-sampling in both modules) (ablation)MPNet DWT_STP+LTR · mask predictionIOU: 0.8829 -> 0.9017Improved
add sparse self-attention branch to dense branch (ablation)VCNet DSA only · YouTube-VOSPSNR: 29.172 -> 29.885Improved
add consistency loss L_c (ablation)VCNet DSA+SSA · YouTube-VOSPSNR: 29.885 -> 30.528Improved
add consistency loss L_c (ablation)VCNet DSA+SSA · YouTube-VOSLPIPS: 0.6891 -> 0.6556Improved
replace MPNet+FGT with BVINetMPNet+FGT · YouTube-VOS blindPSNR: 27.032 -> 30.528Improved
replace MPNet+WaveFormer with BVINetMPNet+WaveFormer · DAVIS blindPSNR: 29.794 -> 30.961Improved
use blind BVINet instead of non-blind E2FGVIE2FGVI · YouTube-VOSPSNR: 30.064 vs 30.528No clear effect
use VCNet instead of WaveFormerWaveFormer · YouTube-VOS non-blindPSNR: 33.264 -> 34.107Improved
use VCNet instead of FGT (FLOPs)FGT · efficiencyFLOPs: 455.91G -> 396.35GImproved

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