Quranic sound recognizer to detect any wrong recitation using Deep Learning technique

Available on telegram with username @the_recitation_bot

Link: https://telegram.me/the_recitation_bot

• Database: Mongo
• Algorithm: Convolutional Neural Network
• Containerization: Docker CE
• Orchestration: Docker Swarm

Don't forget to set secrets

• ../secrets/telegramtoken
• ../secrets/mongouser
• ../secrets/mongopass
• ../secrets/mongohost

• pip install -r requirements (delete tensorflow part if you use tensorflow-gpu)
• apt-get install python-tk
• Install ffmpeg

• Run tensorboard: tensorboard --logdir=/tmp/logs/ --port=80 --host=0.0.0.0
• or on mac: python /Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorboard/main.py --logdir=/tmp/logs/ --port=8080 --host=127.0.0.1
• Run mp3 convertion: ffmpeg -i file.mp3 -acodec pcm_u8 -filter:a loudnorm -ar 22050 -y file.wav
• Run ogg convertion: ffmpeg -i file.ogg -filter:a loudnorm -ar 22050 -y file.wav

• $ git add .
• $ git reset
• $ git stash save --keep-index

It's json structured contains 2 schema

• telegram_id
• username
• current_verse
• current_ayah
• stage_level
• score

• stage_number
• verse_number
• total_ayah

• Ganti max_pad_len