From d1555e45008e50cedbc45ec009d6f0d6ed7df88b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Herv=C3=A9=20Bredin?= Date: Fri, 18 Feb 2022 09:48:34 +0100 Subject: [PATCH] doc: add FAQ section --- README.md | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/README.md b/README.md index 2c20799d..13091374 100644 --- a/README.md +++ b/README.md @@ -70,6 +70,17 @@ pip install https://github.com/pyannote/pyannote-audio/archive/develop.zip - [Speaker verification](tutorials/speaker_verification.ipynb) - Visualization and debugging +## Frequently asked questions + +#### **[Pretrained pipelines](https://huggingface.co/models?other=pyannote-audio-pipeline) do not produce good results on my data. What can I do?** + +1. [Annotate](https://github.com/pyannote/pyannote-audio/blob/develop/tutorials/prodigy.md) dozens of conversations manually and separate them into development and test subsets in [`pyannote.database`](https://github.com/pyannote/pyannote-database#speaker-diarization). +2. [Optimize the hyper-parameters](https://github.com/pyannote/pyannote-audio/blob/develop/tutorials/voice_activity_detection.ipynb) of the pretained pipeline using the development set. If performance is still not good enough, go to step 3. +3. Annotate hundreds of conversations manually and set them up as training subset in `pyannote.database`. +4. [Fine-tune](https://github.com/pyannote/pyannote-audio/blob/develop/tutorials/training_a_model.ipynb) the models (on which the pipeline relies) using the training set. +5. [Optimize the hyper-parameters](https://github.com/pyannote/pyannote-audio/blob/develop/tutorials/voice_activity_detection.ipynb) of the pipeline using the fine-tuned models using the development set. If performance is still not good enough, go back to step 3. + + ## Benchmark The pretrained speaker diarization pipeline with default parameters is expected to be much better in v2.0 than in v1.1: