CrowdSR

This is part of the source code of paper CrowdSR: Enabling High-Quality Video Ingest in Crowdsourced Livecast via Super-Resolution.

It just uses for demonstrating the procedure of the system.

Run

0. Requirements:

   • pytorch
   • aiortc
   • aiohttp
   • requests
   • opencv-contrib-python
   • scikit-image

1. Prepare data

   First, preparing a high resolution video(1920*1080) and a low resolution video(960x540) in work directory.

   Second, put the high resolution frames into a dedicated directory for every other video and put them into a directory. For example, the directory should like below.

   candidates
   |--broadcaster1
      |--0001.png
      |--0002.png
      ....
   |--broadcaster2
   ....
   

2. Start service

   Run the python scripts and add the option on your need.

   python server.py

   python client.py

Citing

TBD