AI Paper Atlas

Research map / Image generation / GANs

High-Quality Sound Separation Across Diverse Categories via Visually-Guided Generative Modeling

Chao Huang, Susan Liang, Yapeng Tian, Anurag Kumar, Chenliang Xu

IJCV 2026 · Image generation › GANs · approach: GAN vocoders and audio synthesis · 6 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 Time-Frequency and Time attention blocks to CA blocks instead of ResNet only (ablation)DAVIS · MUSIC valSDR: 9.03 -> 11.88Improved
add audio-visual Feature Interaction Module instead of plain concatenation (ablation)DAVIS · MUSIC valSDR: 10.85 -> 11.88Improved
use both local and global interaction in FIM instead of global only (ablation)DAVIS · MUSIC valSDR: 11.23 -> 11.88Improved
add silence mask-guided sampling with threshold 0.002 (ablation)DAVIS · MUSIC testSDR: 11.53 -> 11.61Improved
use a high silence threshold 0.01 in mask-guided sampling (ablation)DAVIS · MUSIC testSDR: 11.53 -> 11.15Hurt
train with L1 loss instead of L2 (ablation)DAVIS · MUSIC testSDR: 10.84 -> 11.61Improved
train with L1 loss instead of L2 (ablation)DAVIS · AVE testSDR: 4.53 -> 4.86Improved
replace DDPM with Flow MatchingDAVIS · MUSIC testSDR: 11.61 -> 12.01Improved
replace DDPM with Flow MatchingDAVIS · AVE testSDR: 4.86 -> 5.66Improved
use Flow Matching with 2 Euler steps instead of DDPM with 15 DDIM stepsDAVIS · Inference latency (RTX 4090)latency (s): 1.171 -> 0.156Improved
replace global CLIP embeddings with GroundingDINO localized embeddings at test timeDAVIS-Flow · MUSIC testSDR: 12.01 -> 11.75No clear effect
replace global CLIP embeddings with GroundingDINO localized embeddings at test timeDAVIS-Flow · AVE testSDR: 5.66 -> 5.29No 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