Views
No views yet
damo/speech_campplus-transformer_scl_zh-cn_16k-common for youtube2notion Kaggle startup experiments.configuration.json (453 bytes, sha256 d1aaac1fe468a9efdbde606b69ab527146a651e4a91fb10b49a051f77ae01b2f)campplus_cn_encoder.pt (27280631 bytes, sha256 104c182d5865349a4c9a087e2adfe1b0e27f7e989e9b9ec1a9b9566b624413ef)transformer_backend.pt (10740984 bytes, sha256 dd9fb813a66927413fce473087326ca0c66aa049bc1d4fa3064f188fcb102a97)README.md (3468 bytes, sha256 7881bcf17c2c9cde62bb18d816417cfd25d6c68b7890a09d5fc2e1b6ee8912c3).gitattributes (1414 bytes, sha256 f3d04437c926962476701a7939d22f21400b4eb0bdd31a15f0012576f2d00932)| 测试集 | 时间误差(秒) |
|---|---|
| 合成测试集 | 0.03 |
| 真实测试集 | 0.2 |
1from modelscope.pipelines import pipeline
2scl_pipeline = pipeline(
3 task='speaker-diarization',
4 model='damo/speech_campplus-transformer_scl_zh-cn_16k-common',
5 model_revision='v1.0.0'
6)
7input_wav = 'https://modelscope.cn/api/v1/models/damo/speech_campplus-transformer_scl_zh-cn_16k-common/repo?Revision=master&FilePath=examples/scl_example1.wav'
8result = scl_pipeline(input_wav)
9print(result)
10# 一般情况下,在speaker diarization系统中会获得两人的全局embedding,如果额外输入两人的全局embedding会得到更准确的预测结果,例如:
11# result = scl_pipeline(input_wav, embds=[emb1, emb2])
12# print(result)