/AuViS

An automated highlight generation tool for sports videos

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

AuViS

AuViS (Automated Video Summarization) was developed for the final year project of my Engineering curriculum. The code takes in a video of one of the supported sports videos and generates highlights of the said video. It uses YOLOv3 to detect the location of the scoreboard and a custom trained convolutional neural network as the OCR to extract the score out of the frame.


Getting Started

  • Clone this repository
  • cd AuViS
  • Download the model and JSON files from Release v1.0
  • Install the required modules (pip install -r requirements.txt)
  • Download and install Microsoft Visual C++
  • Copy the desired video into this directory
  • Run Highlights.py

How to Use

  • Run Highlights.py with the desired video in the same directory
  • Select the game
  • Input the name of the video file (with extension)
  • The output gets generated in a sub-directory of the same name as that of the video file
  • You may tweak the lines marked as '#%%' in Highlights.py and Editor.py based on your preference

(Note: As of 2020, this works on latest editions of Bundesliga, English Premier League, Indian Pro Kabaddi, Indian Premier League, ICC Worldcup 2019)


Example Output