|
--- |
|
library_name: transformers |
|
license: mit |
|
language: fr |
|
datasets: |
|
- Cnam-LMSSC/vibravox |
|
metrics: |
|
- per |
|
tags: |
|
- audio |
|
- automatic-speech-recognition |
|
- speech |
|
- phonemize |
|
- phoneme |
|
model-index: |
|
- name: Wav2Vec2-base French finetuned for Speech-to-Phoneme by LMSSC |
|
results: |
|
- task: |
|
name: Speech-to-Phoneme |
|
type: automatic-speech-recognition |
|
dataset: |
|
name: Vibravox["temple_vibration_pickup"] |
|
type: Cnam-LMSSC/vibravox |
|
args: fr |
|
metrics: |
|
- name: Test PER, in-domain training | |
|
type: per |
|
value: 14.2 |
|
--- |
|
|
|
# Model Card |
|
|
|
- **Developed by:** [Cnam-LMSSC](https://huggingface.co/Cnam-LMSSC) |
|
- **Model type:** [Wav2Vec2ForCTC](https://huggingface.co/transformers/v4.9.2/model_doc/wav2vec2.html#transformers.Wav2Vec2ForCTC) |
|
- **Language:** French |
|
- **License:** MIT |
|
- **Finetuned from model:** [facebook/wav2vec2-base-fr-voxpopuli-v2](https://huggingface.co/facebook/wav2vec2-base-fr-voxpopuli-v2) |
|
- **Finetuned dataset:** `temple_vibration_pickup` audio of the `speech_clean` subset of [Cnam-LMSSC/vibravox](https://huggingface.co/datasets/Cnam-LMSSC/vibravox) |
|
- **Samplerate for usage:** 16kHz |
|
|
|
<p align="center"> |
|
<img src="https://cdn-uploads.huggingface.co/production/uploads/6390fc80e6d656eb421bab69/KkZoQQmrn53U6BTLmr0XK.png" /> |
|
</p> |
|
|
|
## Output |
|
|
|
As this model is specifically trained for a speech-to-phoneme task, the output is sequence of [IPA-encoded](https://en.wikipedia.org/wiki/International_Phonetic_Alphabet) words, without punctuation. |
|
If you don't read the phonetic alphabet fluently, you can use this excellent [IPA reader website](http://ipa-reader.xyz) to convert the transcript back to audio synthetic speech in order to check the quality of the phonetic transcription. |
|
|
|
## Link to other phonemizer models trained on other body conducted sensors : |
|
|
|
An entry point to all **phonemizers** (speech-to-phoneme ASR) models trained on different sensor data from the trained on different sensor data from the [Vibravox dataset](https://huggingface.co/datasets/Cnam-LMSSC/vibravox) is available at [https://huggingface.co/Cnam-LMSSC/vibravox_phonemizers](https://huggingface.co/Cnam-LMSSC/vibravox_phonemizers). |
|
|
|
### Disclaimer |
|
Each of these models has been trained for a **specific non-conventional speech sensor** and is intended to be used with **in-domain data**. The only exception is the headset microphone phonemizer, which can certainly be used for many applications using audio data captured by airborne microphones. |
|
|
|
Please be advised that using these models outside their intended sensor data may result in suboptimal performance. |
|
|
|
## Training procedure |
|
|
|
The model has been finetuned for 10 epochs with a constant learning rate of *1e-5*. To reproduce experiment please visit [jhauret/vibravox](https://github.com/jhauret/vibravox). |
|
|
|
## Inference script : |
|
|
|
```python |
|
import torch, torchaudio |
|
from transformers import AutoProcessor, AutoModelForCTC |
|
from datasets import load_dataset |
|
|
|
processor = AutoProcessor.from_pretrained("Cnam-LMSSC/phonemizer_temple_vibration_pickup") |
|
model = AutoModelForCTC.from_pretrained("Cnam-LMSSC/phonemizer_temple_vibration_pickup") |
|
test_dataset = load_dataset("Cnam-LMSSC/vibravox", "speech_clean", split="test", streaming=True) |
|
|
|
audio_48kHz = torch.Tensor(next(iter(test_dataset))["audio.temple_vibration_pickup"]["array"]) |
|
audio_16kHz = torchaudio.functional.resample(audio_48kHz, orig_freq=48_000, new_freq=16_000) |
|
|
|
inputs = processor(audio_16kHz, sampling_rate=16_000, return_tensors="pt") |
|
logits = model(inputs.input_values).logits |
|
predicted_ids = torch.argmax(logits,dim = -1) |
|
transcription = processor.batch_decode(predicted_ids) |
|
|
|
print("Phonetic transcription : ", transcription) |
|
``` |
|
|