Spaces:
Running
Running
| title: README | |
| emoji: 🏃 | |
| colorFrom: indigo | |
| colorTo: purple | |
| sdk: static | |
| pinned: false | |
| [diarizers-community](https://huggingface.co/diarizers-community) aims to promote speaker diarization on the Hugging Face hub. It contains: | |
| - A collection of [multilingual speaker diarization datasets](https://huggingface.co/collections/diarizers-community/speaker-diarization-datasets-66261b8d571552066e003788) that are compatible with the [diarizers](https://github.com/huggingface/diarizers) library. They have been processed using [diarizers scripts](https://github.com/huggingface/diarizers/blob/main/datasets/README.md). | |
| The available datasets are the CallHome (Japanese, Chinese, German, Spanish, English), AMI Corpus (English), Vox-Converse (English) and Simsamu (French). We aim to add more datasets in the future to better support speaker diarization on the Hub. | |
| - A collection of multilingual [fine-tuned segmentation model](https://huggingface.co/collections/diarizers-community/models-66261d0f9277b825c807ff2a) baselines compatible with [pyannote](https://github.com/pyannote/pyannote-audio). | |
| Each model has been fine-tuned on a specific Callhome language subset. They achieve better performances on multilingual data compared to [pyannote](https://github.com/pyannote/pyannote-audio)'s pre-trained [segmentation-3.0](https://huggingface.co/pyannote/segmentation-3.0) model (see benchmark for more details on model performance). | |
| Together with diarizers-community, we release: | |
| - [diarizers](https://github.com/huggingface/diarizers/tree/main), a library for fine-tuning [pyannote](https://github.com/pyannote/pyannote-audio) speaker diarization models using the Hugging Face ecosystem. | |
| - A google colab [notebook](https://colab.research.google.com/github/kamilakesbi/notebooks/blob/main/fine_tune_pyannote.ipynb), with a step-by-step guide on how to use diarizers. | |
| **Benchmark** | |
| | [Callhome](https://huggingface.co/datasets/diarizers-community/callhome) test dataset | Model | DER | False alarm | Missed detection| Confusion | | |
| | ------------------------| ------------- | ------------- | ------------- | --------------- | ------------- | | |
| | Japanese | [Pretrained](https://huggingface.co/pyannote/segmentation-3.0) | 25.44 | **2.30** | 17.45 | 5.69 | | |
| | | [Fine-tuned](https://huggingface.co/diarizers-community/speaker-segmentation-fine-tuned-callhome-jpn) | **18.23** | 6.31 | **6.91** | **5.01** | | |
| | Spanish | [Pretrained](https://huggingface.co/pyannote/segmentation-3.0) | 33.44 | **2.59** | 25.19 | **5.66** | | |
| | | [Fine-tuned](https://huggingface.co/diarizers-community/speaker-segmentation-fine-tuned-callhome-spa) | **25.72** | 6.87 | **12.73** | 6.12 | | |
| | English | [Pretrained](https://huggingface.co/pyannote/segmentation-3.0) | 22.16 | **6.29** | 10.97 | 4.90 | | |
| | | [Fine-tuned](https://huggingface.co/diarizers-community/speaker-segmentation-fine-tuned-callhome-eng) | **18.40** | 7.10 | **6.98** | **4.32** | | |
| | German | [Pretrained](https://huggingface.co/pyannote/segmentation-3.0) | 21.90 | **3.10** | 14.25 | 4.55 | | |
| | | [Fine-tuned](https://huggingface.co/diarizers-community/speaker-segmentation-fine-tuned-callhome-deu) | **16.75** | 5.00 | **7.75** | **4.00** | | |
| | Chinese | [Pretrained](https://huggingface.co/pyannote/segmentation-3.0) | 19.73 | **4.81** | 9.82 | 5.11 | | |
| | | [Fine-tuned](https://huggingface.co/diarizers-community/speaker-segmentation-fine-tuned-callhome-zho) | **15.95** | 5.04 | **7.24** | **3.68** | | |
| Results are in %. They have been obtained using the [test script](https://github.com/huggingface/diarizers/blob/main/test_segmentation.py) from diarizers. | |