AI Paper Atlas

Research map / Video / Video segmentation and tracking

Joint-Modal Label Denoising for Weakly-Supervised Audio-Visual Video Parsing

Haoyue Cheng, Zhaoyang Liu, Hang Zhou, Chen Qian, Wayne Wu, Limin Wang

ECCV 2022 · Video › Video segmentation and tracking · approach: Self-supervised video representation learning · 36 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 estimated noise ratios instead of constant 0.5 (ablation)JoMoLD · LLP segment-levelVisual F: 58.3 -> 63.8Improved
use estimated noise ratios instead of constant 0.5 (ablation)JoMoLD · LLP segment-levelType@AV F: 56.6 -> 60.8Improved
remove cross-modal attention from noise estimator (ablation)JoMoLD · LLP segment-levelVisual F: 55.9 -> 63.8Improved
skip cross-modal attention when computing forward loss (ablation)JoMoLD · LLP segment-levelType@AV F: 58.9 -> 60.8Improved
use joint-modal instead of intra-modal label denoising (ablation)JoMoLD · LLP segment-levelVisual F: 58.5 -> 63.8Improved
use joint-modal instead of intra-modal label denoising (ablation)JoMoLD · LLP segment-levelType@AV F: 56.6 -> 60.8Improved
denoise both modalities instead of audio only (ablation)JoMoLD · LLP segment-levelVisual F: 53.2 -> 63.8Improved
denoise both modalities instead of visual only (ablation)JoMoLD · LLP segment-levelType@AV F: 60.0 -> 60.8Improved
add warm-up of noise ratios (0.9) (ablation)JoMoLD · LLP segment-levelType@AV F: 60.1 -> 60.8Improved
vary thresholds theta_a and theta_v within range (ablation)JoMoLD · LLP segment-levelType@AV F: 60.5 - 60.8 near optimumNo clear effect
batch size 128 instead of 32 (ablation)JoMoLD · LLP segment-levelType@AV F: 60.3 -> 60.8Improved
apply JoMoLD to AVSlowFast backboneAVSlowFast · LLP segment-levelVisual F: 50.8 -> 60.1Improved

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