Downloads · 30 days
7
19% of all-time downloads
3BDOAi3/facebookwav2vec2-xls-r-300m-finetuned-with-MTC-Dataset
facebookwav2vec2-xls-r-300m-finetuned-with-MTC-Dataset is a automatic speech recognition model from 3BDOAi3. Use it when you need speech turned into text. It is set up for transformers.
This project aims to develop a robust Speech-to-Text (STT) system specifically designed for transcribing Egyptian dialect speech into text. The system leverages the pre-trained model facebookwav2vec2-xls-r-300m and em…
Downloads · 30 days
7
19% of all-time downloads
All-time downloads
36
Public
Repo size
2.5 GB
Likes
0
Public
Click a slice to open those files.
.bin1.3 GB · 100%
From the Hugging Face model README
This project aims to develop a robust Speech-to-Text (STT) system specifically designed for transcribing Egyptian dialect speech into text. The system leverages the pre-trained model facebookwav2vec2-xls-r-300m and employs phased training with a large dataset to achieve accurate transcription.
The primary objective is to enhance speech recognition capabilities for the Egyptian dialect through fine-tuning of the Wav2Vec2 model, ensuring high accuracy and reliability.
datasets, torchaudio, transformers, pandas, numpy, librosa, torch, etc.Download the codes
Install dependencies using pip
pip install -r requirements.txt
Download and extract the dataset
train and adapt).Data and Resources Dataset Information MTC-ASR-Dataset-16K Arabic Speech Recognition for Egyptian Dialects using a data resource consisting of 100 hrs recorded Egyptian dialect samples.
Duration: 3 hours Number of files: 1726 audio 16KHz WAV files Environment: Clean and Noisy
train/: Contains the training audio data (WAV files). adapt/: Contains the adaptation audio data (WAV files). train.csv: CSV file with wav_id and transcription columns for training data. adapt.csv: CSV file with wav_id and transcription columns for adaptation data.
Download the dataset from the provided link and extract it into the respective directories (train and adapt).
Preprocess_&_Prepare_Data.ipynb on Kaggle TPU.Train.ipynb on your local machine, using phased training due to the large dataset.Test.ipynb on your local machine.Preprocess_&_Prepare_Data.ipynb on Kaggle.Train.ipynb on your local machine.Test.ipynb on your local machine.