LoCoNet: Long-Short Context Network for Active Speaker Detection

Open paper

Benchmark	Model	Rank	Results
audio-visual-active-speaker-detection-on-ava-activespeaker	LoCoNet	#3	validation mean average precision: 95.2%