Hi everyone. 
We are a team of company  "Constructor" and we develop cutting edge cartographic systems. Recently the amount of image data skyrocketed so we want to give our users the ability to tie images to places automatically. For that, we want to use AI/ML technologies and we have a cool task for you.

https://cloud.mail.ru/public/pHbC/4r7Z58m6f/

There are three collections of datasets and in each you have:
Image from the original camera with no position information and set of images made from different points of view near this original camera with position information (list_files_info.