Eqraa is a project launched to aid in the development of recitation and memorization of the nobel Quran. This project leverages machine learning techniques imporves pronunciation, and provides feedback for better interactive experience.
- Speech Recognition: Analyzes spoken words to provide feedback and ensure accurate pronunciation.
- Keyword Spotting System: Check how close is the spoken word to the optimal pronunciation.
- Interactive Reading: Provides real-time assistance and feedback during reading.
To install and run Eqraa locally, follow these steps:
- Clone the repository:
git clone https://github.com/AHedya/Eqraa.git cd Eqraa - Initialize Anaconca env:
conda env create -f environment.yml conda activate torch-gpu // or whatever name you choose for the environment - Run the server:
fastapi dev server.py
make sure server running at (http://127.0.0.1:8000)
- Try client side:
Open
client.ipynb, and now you can try the project yourself.
You can find demo of the project running at project demo.
Training results, and additional details at implemntation docs, page 10 to the end
All scripts and trained models variations are at scripts
-
To accomplish this project we used our from-scratch deep learning models, packaged the into singleton classes, And also we used some pretrained models such as tarteel-base,tarteel-tiny and Quran syllables
-
Pre-trained models are ready to be utilized on the whole Moshaf, although we only used them on AlFatiha, while our from-scratch models can only be applied on AlFatiha.
Second version of our data set is available on Kaggle. Make sure You're viewing second version.