As and example, we used (nvidia-)docker images to make our object detection results reproducible: https://github.com/umautobots/driving-in-the-matrix https://g

As and example, we used (nvidia-)docker images to make our object detection results reproducible: https://github.com/umautobots/driving-in-the-matrix https://github.com/umautobots/driving-in-the-matrix docker has been great for this in the lab: only one person now goes through the pain to get latest framework + hack + model tweak work and then the rest can reuse.