PATS dataset consists of a diverse and large amount of aligned pose, audio and transcripts. With this dataset, we hope to provide a benchmark that would help develop technologies for virtual agents which generate natural and relevant gestures.
9 PAPERS • NO BENCHMARKS YET