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

Successful active speaker detection requires a threestage pipeline: (i) audio-visual encoding for all speakers in the clip, (ii) inter-speaker relation modeling between a reference speaker and the background speakers within each frame, and (iii) temporal modeling for the reference speaker. Each stage of this pipeline plays an important role for the final performance of the created architecture. Based on a series of controlled experiments, this work presents several practical guidelines for audio-visual active speaker detection. Correspondingly, we present a new architecture called ASDNet, which achieves a new state-of-the-art on the AVA-ActiveSpeaker dataset with a mAP of 93.5% outperforming the second best with a large margin of 4.7%. Our code and pretrained models are publicly available 1 . Recently, the AVA-ActiveSpeaker dataset [42] provided the first large-scale standard benchmark for audio-visual active speaker detection in the wild. Recent research [1, 32] indicates that active speaker detection in the wild requires (i) integration of audio-visual information for each speaker, (ii) contextual information that captures inter-speaker relationships, and (iii) temporal modeling to exploit long term relationships in natural conversation. In this paper, we con-