bhuiyans-dataset-builder

Dataset-builder

A video processing tool for visual data collection, end-to-end preprocessing, ready-to-go for model training.

Input

A video or a set of videos. Currently supports .mp4 only.

Features

• Audio extraction :v:
  • python src/app.py --fileName 001-01.mp4 --extractAudio yes
• Audio detection :v:
  • python src/app.py --fileName 001-01.wav --detectAudio yes
  • To show plot: python src/app.py --fileName 001-01.wav --detectAudio yes --plot true
• Audio-video split :v:
  • To separate audio: python src/app.py --fileName 001-01.wav --separateAudio yes
  • To separate video: python src/app.py --fileName 001-01.mp4 --separateVideo yes
• Audio-video merge :v:
  • python src/app.py --fileName 001-01.mp4 --mergeAudioVideo yes
• Video player :v:
  • Play the separated videos to see all is good: python src/app.py --fileName 001-01.mp4 --playAll
• Video compression :v:
  • python src/app.py --fileName 001-01.mp4 --compressBySize 2
• Face detection and region extraction :v:
  • python src/app.py --fileName 001-01.mp4 --detectLip yes --speaker 001
• Video data processing utilities :v:

How to use it?

• Clone the repository: https://github.com/MasumBhuiyan/visual-data-manager.git
• Open terminal and cd to the directory where requirements.txt is located.
• Create, activate, and install packages in a virtual environment
  • pip install virtualenv
  • virtualenv env
  • source env/bin/activate
  • pip install -r requirements.txt
• To split a video run: will be updated