AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

How To Design a Three-Stage Architecture for Audio-Visual Active Speaker Detection in the Wild

Okan Köpüklü, Maja Taseska, Gerhard Rigoll

ICCV 2021 · Vision-language and multimodal › Cross-modal learning · approach: Audio-visual source localization · 54 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
replace 2D-ResNet-18 video backbone with 3D-ResNet-18 (ablation)audio-visual encoder · AVA-ActiveSpeaker val (encoder only, 8 frames)mAP (%): 79.0 -> 83.9Improved
replace MFCC 2D-ResNet-18 audio backbone with SincDSNet (ablation)audio-visual encoder · AVA-ActiveSpeaker val (encoder only, 8 frames)mAP (%): 79.0 -> 80.8Improved
use both 3D-ResNet-18 and SincDSNet (ablation)audio-visual encoder · AVA-ActiveSpeaker val (encoder only, 8 frames)mAP (%): 79.0 -> 86.1Improved
use 3D-ResNeXt-101 instead of 3D-ResNet-18 (ablation)video backbone · 8-frame clipsmAP (%): 86.1 -> 86.7Improved
increase clip length from 8 to 16 frames (ablation)3D-ResNeXt-101 encoder · AVA-ActiveSpeaker valmAP (%): 86.7 -> 88.9Improved
increase clip length from 16 to 32 frames (ablation)3D-ResNeXt-101 encoder · AVA-ActiveSpeaker valmAP (%): 88.9 -> 88.9No clear effect
add temporal modeling to encoder (ablation)ASDNet · AVA-ActiveSpeaker val (component analysis)mAP (%): 88.9 -> 92.6Improved
add ISRM to encoder (ablation)ASDNet · AVA-ActiveSpeaker val (component analysis)mAP (%): 88.9 -> 89.6Improved
add ISRM on top of encoder plus temporal modeling (ablation)ASDNet · AVA-ActiveSpeaker val (component analysis)mAP (%): 92.6 -> 93.5Improved
use NonLocal instead of ISRM with temporal model (ablation)ASDNet · AVA-ActiveSpeaker val (ISRM comparison)mAP (%): 92.8 -> 93.4Improved
use background speaker features from 9 neighbouring frames (ablation)ISRM · AVA-ActiveSpeaker valmAP (%): 93.4 -> 93.5Improved
use bidirectional GRU instead of unidirectional GRU (ablation)temporal model · AVA-ActiveSpeaker val (temporal model comparison)mAP (%): 92.8 -> 93.5Improved

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