AI Paper Atlas

Research map / Video / Video-language understanding

Beyond Simple Edits: Composed Video Retrieval with Dense Modifications

Omkar Thawakar, Dmitry Demidov, Ritesh Thawkar, Rao Muhammad Anwer, Mubarak Shah, Fahad Shahbaz Khan, Salman Khan

ICCV 2025 · Video › Video-language understanding · approach: Text-video retrieval · 5 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 dense descriptions and dense modification text in training (ablation)Ours · Dense-WebVid-CoVR testR@1: 60.4 -> 71.2Improved
add dense modification text (descriptions kept WebVid-CoVR) (ablation)Ours · Dense-WebVid-CoVR testR@1: 60.4 -> 61.2Improved
add dense descriptions (modification text kept WebVid-CoVR) (ablation)Ours · Dense-WebVid-CoVR testR@1: 60.4 -> 63.8Improved
use dense modification text at inference (ablation)Ours · Dense-WebVid-CoVR testR@1: 68.88 -> 71.26Improved
use dense video descriptions at inference (ablation)Ours · Dense-WebVid-CoVR testR@1: 66.08 -> 71.26Improved
use unified weighted-mean fusion instead of addition (ablation)Ours · Dense-WebVid-CoVR testR@1: 69.72 -> 71.26Improved
use unified weighted-mean fusion instead of cross-attention for video and description (ablation)Ours · Dense-WebVid-CoVR testR@1: 70.13 -> 71.26Improved
replace pairwise fusion with unified single grounding encoder fusion (no dense training)Thawakar et al. · Dense-WebVid-CoVR test (CA, no dense training)R@1: 39.20 -> 48.08Improved
replace pairwise fusion with unified fusion (trained on dense data)Thawakar et al. · Dense-WebVid-CoVR testR@1: 67.86 -> 71.26Improved
train on Dense-WebVid-CoVR rather than notOurs · Dense-WebVid-CoVR test (Text only)R@1: 24.12 -> 50.12Improved
unified fusion vs pairwise fusion inference speedThawakar et al. · Dense-WebVid-CoVRinference speed: 3x fasterImproved
apply unified fusion to composed image retrieval, training on CIRRThawakar et al. · CIRR testR@1: 51.03 -> 56.30Improved

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