Benchmark-ready packaging of the CFAD (Chinese Fake Audio Detection) clean test
set (arXiv 2207.12308), for speech anti-spoofing and
synthetic / deepfake voice detection on Mandarin Chinese speech.
CFAD is a large-scale Chinese fake-audio detection corpus. This repo packages the clean
version's two test partitions:
test_seen — spoof systems and real corpora also present in the train/dev splits.
test_unseen — spoof systems and real corpora held out… See the full description on the dataset page:
https://huggingface.co/datasets/SpeechAntiSpoofingBenchmarks/CFAD.
