The logos have been resized to a uniform shape of 70x70 to make it less demanding on computational resources and model complexity when training. Though the original files can be
found on one of my github repos, the link can be found below.
GitHub Repo.
The code used to mine and process the data can also be found on the repo.