AI Paper Atlas

Research map / Video / Video segmentation and tracking

Exploring Cross-Video and Cross-Modality Signals for Weakly-Supervised Audio-Visual Video Parsing

Yan-Bo Lin, Hung-Yu Tseng, Hsin-Ying Lee, Yen-Yu Lin, Ming-Hsuan Yang

NeurIPS 2021 · Video › Video segmentation and tracking · approach: Self-supervised video representation learning · 28 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 cross-video shared semantics (CM-S) (ablation)HAN · LLP testvisual segment F-score: 52.9 -> 58.3Improved
add cross-video shared semantics (CM-S) (ablation)HAN · LLP testaudio segment F-score: 60.1 -> 58.1Hurt
add cross-modality co-occurrence (CM-Co) (ablation)HAN · LLP testEvent@AV segment F-score: 55.4 -> 57.4Improved
add CM-S (ablation)MA* · LLP testvisual segment F-score: 57.5 -> 60.7Improved
add CM-Co (ablation)MA* · LLP testEvent@AV segment F-score: 56.6 -> 59.1Improved
add both CM-S and CM-Co (ablation)MA* · LLP testvisual segment F-score: 57.5 -> 63.5Improved
add both CM-S and CM-Co (ablation)HAN · LLP testType@AV segment F-score: 54.0 -> 57.5Improved
co-occurrence on visual classes only instead of cross-modality co-attention (ablation)MA* · LLP testaudio segment F-score: 60.5 -> 46.8Hurt
add cross-modality co-attention on top of within-modality self-attention (ablation)MA* · LLP testEvent@AV segment F-score: 56.7 -> 59.1Improved
co-occurrence on audio classes only (ablation)MA* · LLP testaudio segment F-score: 59.8 -> 60.7Improved
add more transformer layers to HAN instead of CM-Co (ablation)HAN · LLP testEvent@AV segment F-score: 55.4 -> 54.9No clear effect
add CM-Co module vs adding a layer (ablation)HAN · LLP testEvent@AV segment F-score: 54.9 -> 57.4Improved

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