AI Paper Atlas

Research map / Video / Video-language understanding

Multi-modal Grouping Network for Weakly-Supervised Audio-Visual Video Parsing

Shentong Mo, Yapeng Tian

NeurIPS 2022 · Video › Video-language understanding · approach: Dense video captioning · 4 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 class-aware unimodal grouping (CUG) to baseline (ablation)HAN baseline · LLP segment-levelF-score Visual: 52.9 -> 55.3Improved
Add CUG to baseline (ablation)HAN baseline · LLP segment-levelF-score Audio: 60.1 -> 58.9Hurt
Add CUG to baseline (ablation)HAN baseline · LLP segment-levelF-score Event@AV: 55.4 -> 55.9Improved
Add modality-aware cross-modal grouping (MCG) on top of CUG (ablation)MGN · LLP segment-levelF-score Audio-Visual: 49.8 -> 50.6Improved
Add MCG on top of CUG (ablation)MGN · LLP segment-levelF-score Type@AV: 54.7 -> 55.6Improved
Add MCG on top of CUG (ablation)MGN · LLP segment-levelF-score Event@AV: 55.9 -> 57.2Improved
Use MGN instead of MA (w C) on audio-visual contrastive learningMGN (w C) · LLP segment-levelF-score Visual: 53.1 -> 56.7Improved
Use MGN instead of MA with label refinementMGN (w R) · LLP segment-levelF-score Visual: 57.5 -> 60.6Improved
Use MGN instead of MA with label refinementMGN (w R) · LLP event-levelF-score Visual: 54.4 -> 58.4Improved
Use MGN (w C+R) instead of MA (w C+R)MGN (w C+R) · LLP event-levelF-score Event@AV: 50.6 -> 49.9Hurt
Use MGN instead of HANMGN · LLP segment-levelF-score Audio: 60.1 -> 60.7No clear effect
Use MGN instead of HANMGN · LLP event-levelF-score Audio: 51.3 -> 51.0No clear effect

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